Rajeev Acharya, Dmitry A. Abanin and Laleh Aghababaie-Beni
Google Quantum AI demonstrated a superconducting surface-code memory whose logical error rate decreases as the code distance grows, crossing below the fault-tolerance threshold. Scaling from distance-3 to distance-5 to distance-7 codes, the logical qubit's error per cycle was suppressed by roughly a factor of two per increment, showing exponential error suppression. This provides experimental evidence that the surface code can reach the regime needed for scalable fault-tolerant quantum computing.
Using JWST/NIRSpec observations from the JWST Advanced Deep Extragalactic Survey (JADES), the authors obtained spectroscopic confirmation of two unusually luminous galaxies, JADES-GS-z14-0 and JADES-GS-z14-1, at redshifts of about 14. These are among the most distant galaxies ever spectroscopically confirmed, existing roughly 290–300 million years after the Big Bang. Their brightness challenges pre-JWST models of how rapidly luminous galaxies could form in the early universe.
This paper introduced AlphaFold 3, a unified deep learning model that predicts the joint structure of complexes containing proteins, nucleic acids, small-molecule ligands, ions, and modified residues. It replaces much of the prior architecture with a diffusion-based module that directly generates atomic coordinates. The model achieved substantially improved accuracy over specialized tools across many interaction types, including protein-ligand and protein-nucleic acid complexes.
Dolev Bluvstein, Simon J. Evered and Alexandra A. Geim
The authors demonstrated a programmable quantum processor using reconfigurable arrays of neutral atoms that operates on encoded logical qubits rather than physical ones. They ran error-correcting codes, performed logical entangling operations and algorithms on dozens of logical qubits, and showed that increasing code distance improved logical performance. The work is a key step toward fault-tolerant quantum computation with atom arrays.
This paper introduced RFdiffusion, a generative diffusion model built on the RoseTTAFold network for de novo protein design. It enables a range of design tasks, including unconditional generation, symmetric oligomer design, functional motif scaffolding, and binder design. Many designs were experimentally validated, with solved structures closely matching the intended models.
Gabriella Agazie, Akash Anumarlapudi and Anne M. Archibald
Using 15 years of pulsar timing data from a 67-pulsar array, the NANOGrav Collaboration reported evidence for a stochastic gravitational-wave background at nanohertz frequencies. The team detected the characteristic spatial cross-correlation (Hellings-Downs) signature expected for such a background across pulsar pairs, at a significance of roughly 3 to 4 sigma. The signal is consistent with a population of inspiraling supermassive black hole binaries, though exotic cosmological sources are not excluded.
The Human Pangenome Reference Consortium presents a first draft human pangenome built from 47 phased, diploid genome assemblies of genetically diverse individuals. The assemblies cover more than 99% of the expected sequence per genome at over 99% base-level and structural accuracy, and are combined into a graph-based reference. Relative to GRCh38, the pangenome adds about 119 million base pairs of euchromatic polymorphic sequence and 1,115 gene duplications, improving representation of variation at structurally complex loci.
This paper introduces the Segment Anything project: a promptable image segmentation task, the Segment Anything Model (SAM), and the SA-1B dataset. SAM combines an image encoder, a flexible prompt encoder (points, boxes, masks, text), and a fast mask decoder to produce valid segmentation masks from arbitrary prompts. Trained on over 1 billion masks across 11 million images, SAM shows strong zero-shot transfer to many segmentation tasks without additional training.
This technical report describes GPT-4, a large-scale multimodal Transformer model that accepts image and text inputs and produces text outputs. The report emphasizes that GPT-4 achieves human-level performance on a range of professional and academic benchmarks, and details infrastructure and optimization methods that allowed performance to be predicted from much smaller models. For competitive and safety reasons, the report withholds architecture, dataset, and training details.
The paper presents LLaMA, a family of foundation language models ranging from 7B to 65B parameters trained exclusively on publicly available datasets. It argues that strong performance can be reached without proprietary data and at smaller parameter counts than prior models. LLaMA-13B outperforms the much larger GPT-3 175B on most benchmarks, and LLaMA-65B is competitive with the best contemporary models such as Chinchilla-70B and PaLM-540B.
Raj Chetty, Matthew O. Jackson, Theresa Kuchler and Johannes Stroebel
Using data on 21 billion Facebook friendships, the authors construct ZIP-code-level measures of three distinct forms of social capital: cross-class connectedness, social cohesion, and civic engagement. They show these measures vary widely across areas and are only weakly correlated with one another. They find that economic connectedness between low- and high-income people is strongly associated with upward income mobility, more so than other forms of social capital.
This SH0ES Team paper presents a refined local measurement of the Hubble constant using Hubble Space Telescope observations of Cepheid variables calibrating Type Ia supernovae. By expanding and improving the distance-ladder sample, the authors achieve roughly 1 km/s/Mpc total uncertainty. The result reinforces the significant tension with the Hubble constant inferred from the early-universe CMB.
Edward J. Hu, Yelong Shen, Phillip Wallis, Zeyuan Allen-Zhu, Yuanzhi Li, Shean Wang, et al.
The paper introduces LoRA, a parameter-efficient fine-tuning method that keeps the pretrained model weights frozen and instead learns small trainable low-rank decomposition matrices injected into the Transformer layers. This drastically cuts the number of trainable parameters and optimizer memory needed to adapt very large models to downstream tasks. The authors show LoRA matches or exceeds full fine-tuning quality across several models including GPT-3 175B while adding no extra inference latency.
The Telomere-to-Telomere (T2T) Consortium reports T2T-CHM13, the first essentially gapless assembly of a human genome (all chromosomes except Y), totaling about 3.055 billion base pairs. The assembly resolves previously unfinished heterochromatic and repetitive regions, including centromeric satellite arrays, segmental duplications, and the short arms of the acrocentric chromosomes. It adds nearly 200 million base pairs of new sequence and corrects errors in prior reference assemblies.
Jordan Hoffmann, Sebastian Borgeaud and Arthur Mensch
This paper (the 'Chinchilla' paper) investigates the compute-optimal trade-off between model size and training-token count for large language models. By training over 400 models from 70M to 16B parameters on 5B to 500B tokens, the authors find that model size and training data should be scaled in roughly equal proportion—implying that prior large models were significantly undertrained. Their 70B-parameter Chinchilla model, trained on far more data under the same compute budget as Gopher, outperformed much larger models.
Long Ouyang, Jeff Wu, Xu Jiang, Diogo Almeida, Carroll L. Wainwright, Pamela Mishkin, et al.
The paper (InstructGPT) shows how to align language models with user intent by fine-tuning GPT-3 on human-written demonstrations and then optimizing against a learned reward model with reinforcement learning from human feedback (RLHF). Human evaluators preferred outputs from a 1.3B-parameter InstructGPT model over the 175B GPT-3 model, despite the large size difference. The approach improves truthfulness and reduces toxic generations while causing only minimal regressions on standard NLP benchmarks.
Jason Wei, Xuezhi Wang, Dale Schuurmans, Maarten Bosma, Brian Ichter, Fei Xia, et al.
The paper shows that prompting a large language model with a few exemplars that include intermediate reasoning steps (a 'chain of thought') substantially improves its ability to solve multi-step reasoning problems. This reasoning ability emerges only in sufficiently large models and requires no fine-tuning. Across arithmetic, commonsense, and symbolic reasoning tasks, chain-of-thought prompting produces large gains, including a new state of the art on the GSM8K math word-problem benchmark.
Researchers at the National Ignition Facility reported reaching a burning-plasma regime in laser-driven inertial confinement fusion, where alpha-particle self-heating from fusion reactions becomes the dominant heating mechanism in the deuterium-tritium fuel. Using improved hohlraum and capsule designs, several experiments crossed into this state, producing substantially higher fusion energy yields. The work marked a key milestone on the path toward ignition.
Thibaut Lamadon, Magne Mogstad and Bradley Setzler
The authors build and estimate an equilibrium model of imperfect competition in the US labor market using linked employer-employee administrative data, combining firm-level productivity shocks with worker mobility. They quantify the degree of employer wage-setting power (monopsony), how firms share rents with workers, and the role of compensating differentials for non-wage amenities. They find that firms have substantial market power and pass through only part of productivity gains to wages, while amenities matter for worker sorting.
Robin Rombach, Andreas Blattmann, Dominik Lorenz, Patrick Esser and Björn Ommer
The paper proposes latent diffusion models (LDMs), which apply the diffusion process in the compressed latent space of a pretrained autoencoder rather than directly in pixel space, greatly reducing compute. A cross-attention conditioning mechanism enables flexible inputs such as text and bounding boxes for tasks including text-to-image generation, inpainting, and super-resolution. LDMs achieve strong or state-of-the-art results across these tasks while being far more efficient to train and sample, and this architecture underlies Stable Diffusion.
Using a superconducting quantum processor, the team prepared the ground state of the toric code, a paradigmatic topologically ordered state, on a 2D lattice of qubits. They demonstrated the topological nature of the state by creating and braiding anyonic excitations and measuring nontrivial topological entanglement entropy. The work showed that programmable quantum hardware can realize and probe states with intrinsic topological order.
The paper develops a framework for difference-in-differences designs in which units adopt treatment at different times across multiple periods. It defines group-time average treatment effects and shows how to identify and estimate them under conditional parallel trends, then aggregate them into interpretable summary parameters. The authors provide valid simultaneous inference and apply the methods to estimating the effect of minimum wage increases on teen employment.
Martin S. Eichenbaum, Sergio Rebelo and Mathias Trabandt
The authors extend the canonical SIR epidemiological model by embedding it in a macroeconomic framework where people's consumption and labor decisions affect the spread of infection. They show that the epidemic causes a sharp recession because infected and susceptible agents cut activity to reduce contagion. A key tension emerges: the competitive equilibrium worsens the epidemic because individuals do not internalize the infection externality, and well-designed containment policy can mitigate the death toll though it deepens the short-run economic contraction.
This paper introduces Enformer, a transformer-based deep learning model that predicts gene expression and chromatin states directly from DNA sequence by integrating regulatory information from up to ~100 kb away. By using self-attention to capture long-range interactions, it substantially improves prediction accuracy over prior convolutional models. The approach also improves prediction of the effects of non-coding genetic variants on expression.
Kathryn Tunyasuvunakool, John Jumper and Demis Hassabis
This companion paper applied AlphaFold to predict structures for nearly the entire human proteome and 20 other key organisms, producing a large public database of predicted models. It assessed coverage and confidence across the human proteome, showing that a substantial fraction of residues could be modeled with high or very high confidence. The work created the AlphaFold Protein Structure Database, greatly expanding structural coverage beyond experimentally determined structures.
John Jumper, Richard Evans, Alexander Pritzel, David Silver, Oriol Vinyals and Demis Hassabis
The paper introduces AlphaFold2, a deep-learning system that predicts three-dimensional protein structures directly from amino-acid sequence with near-experimental accuracy. It combines a novel attention-based Evoformer over multiple sequence alignments and pairwise representations with an end-to-end structure module that produces atomic coordinates. AlphaFold won the CASP14 assessment by a wide margin, delivering atomic-level accuracy for the majority of targets.
This paper presented RoseTTAFold, a three-track neural network that simultaneously processes one-dimensional sequence, two-dimensional residue-pair distances, and three-dimensional atomic coordinate information, with information flowing between the tracks. The method achieved protein structure prediction accuracy approaching that of AlphaFold2 while being more computationally efficient. It also demonstrated rapid generation of accurate models for protein-protein complexes.
Alexey Dosovitskiy, Lucas Beyer, Alexander Kolesnikov and Neil Houlsby
This paper introduced the Vision Transformer (ViT), applying a standard Transformer encoder directly to sequences of image patches treated as tokens, with minimal vision-specific inductive biases. When pre-trained on large datasets and transferred to downstream tasks, ViT matched or exceeded state-of-the-art convolutional networks while requiring fewer computational resources to train. It demonstrated that convolutions are not necessary for strong image recognition at scale.
The Fermilab Muon g-2 Collaboration reported its first measurement of the positive muon's anomalous magnetic moment with 0.46 ppm precision. The result was consistent with the earlier Brookhaven measurement, and the combined experimental value showed a tension of about 4.2 sigma with the Standard Model theoretical prediction. This hinted at possible physics beyond the Standard Model.
Alec Radford, Jong Wook Kim, Chris Hallacy, Aditya Ramesh, Gabriel Goh, Sandhini Agarwal, et al.
The paper presents CLIP, which learns visual representations by contrastively matching images to their natural-language captions over a 400-million-pair web dataset. The pretrained model can be applied zero-shot to many downstream vision tasks by framing class labels as text prompts, without task-specific fine-tuning. It matches the accuracy of a supervised ImageNet ResNet-50 zero-shot and transfers robustly across a broad benchmark suite.
Francisco J. Buera, Joseph P. Kaboski and Yongseok Shin
The paper builds a general-equilibrium model with heterogeneous agents and financial frictions to evaluate the economy-wide effects of large-scale microfinance programs. It finds that while microfinance raises the incomes of marginal entrepreneurs, partial-equilibrium estimates overstate aggregate gains because general-equilibrium wage and capital responses redistribute benefits. The aggregate effects on output and total factor productivity are modest, but the welfare and distributional consequences can be substantial, especially for poorer households.
Fernando P. Polack, Stephen J. Thomas, Nicholas Kitchin, Judith Absalon, Alejandra Gurtman, Stephen Lockhart, et al.
This multinational, randomized, placebo-controlled phase 2/3 trial evaluated the safety and efficacy of two 30-µg doses of the BNT162b2 lipid-nanoparticle mRNA vaccine, given 21 days apart, in 43,548 participants 16 years of age or older. Among those without prior SARS-CoV-2 infection, the vaccine conferred 95% protection against laboratory-confirmed Covid-19 beginning at least 7 days after the second dose. The safety profile over a median of about two months was characterized by short-lived, mostly mild-to-moderate reactions, with a low incidence of serious adverse events comparable to placebo.
The team demonstrates Gaussian boson sampling with a photonic quantum computer named Jiuzhang, detecting up to 76 output photons from a 100-mode interferometer. The resulting output distribution spans an enormous Hilbert space, and they estimate the sampling task would take classical supercomputers astronomically longer, providing a photonics-based demonstration of quantum computational advantage. This complemented the earlier superconducting-qubit supremacy result with a distinct physical platform.
This paper presented GPT-3, an autoregressive language model with 175 billion parameters, and studied its ability to perform tasks from natural-language descriptions and a few examples without gradient updates (in-context learning). Scaling the model dramatically improved few-shot performance across many NLP benchmarks, sometimes approaching fine-tuned systems. The authors also examined limitations, data contamination, and broader societal impacts of large language models.
Longxing Cao, Inna Goreshnik, Brian Coventry and David Baker
The authors used computational de novo protein design to create small, stable miniproteins that bind the SARS-CoV-2 spike receptor-binding domain and block its interaction with ACE2. Two design strategies were used: incorporating the ACE2 helix into a designed scaffold, and building entirely new binders against the RBD. The best designs bound with picomolar affinity and neutralized the virus, with cryo-EM confirming the binding modes matched the computational models.
This paper presents the cosmological parameter constraints from the final (2018) Planck satellite analysis of cosmic microwave background temperature and polarization anisotropies plus lensing. The data are well fit by a six-parameter flat Lambda-CDM model, yielding precise values for the matter density, baryon density, and other parameters. The inferred Hubble constant from the CMB is in tension with local distance-ladder measurements.
The paper shows that linear regressions with group and period fixed effects estimate a weighted sum of treatment effects across cells where some weights can be negative. As a consequence the coefficient can be negative even when every cell-level effect is positive, undermining its interpretation as an average treatment effect. The authors propose an alternative estimator (DID_M) that is robust to heterogeneous effects and assess the empirical relevance of the negative-weighting problem.
The paper introduces denoising diffusion probabilistic models (DDPMs), a class of latent-variable generative models trained to reverse a fixed Gaussian noising process. It establishes a connection between diffusion models and denoising score matching with Langevin dynamics, and proposes a simplified, reweighted training objective. The resulting models produce high-quality image samples, achieving competitive log-likelihoods and a strong FID on CIFAR-10.
The authors study how the spread of industrial robots affected US local labor markets between 1990 and 2007, using variation in industry robot adoption combined with regional industry composition. They estimate the effect on employment and wages in commuting zones more exposed to robotization. They find that each additional robot per thousand workers reduced the employment-to-population ratio and wages in the affected local labor markets, with the largest negative effects on routine manual occupations.
Raj Chetty, Nathaniel Hendren, Maggie R. Jones and Sonya R. Porter
Using de-identified longitudinal data covering nearly the entire U.S. population, the authors study racial and ethnic differences in intergenerational income mobility. They find persistent black-white income gaps driven primarily by differences among men, while black and white women have similar outcomes conditional on parental income. They show neighborhood and family-background factors explain little of the gap, which persists even within the same neighborhoods.
David Autor, David Dorn, Lawrence F. Katz, Christina Patterson and John Van Reenen
The authors document the decline in the labor share of income across US industries and propose a 'superstar firm' explanation: industries are increasingly dominated by highly productive, low-labor-share firms. Using US Economic Census data and cross-country evidence, they show that reallocation of economic activity toward these firms, rather than declines within typical firms, drives the aggregate fall in the labor share. Industries with rising concentration show the largest labor-share declines.
Using firm-level data for the U.S. economy since 1955, the authors estimate price-cost markups and document a sharp rise in aggregate market power beginning around 1980. They argue this increase in markups can account for several secular macroeconomic trends, including the declining labor and capital shares and reduced labor-market dynamism.
David E. Gordon, Gwendolyn M. Jang and Mehdi Bouhaddou
The authors expressed 26 of the 29 SARS-CoV-2 proteins in human cells and used affinity-purification mass spectrometry to map 332 high-confidence virus-human protein-protein interactions. They then identified 69 existing drugs and compounds that target the human proteins in this interactome and tested several for antiviral activity. Two pharmacological classes (mRNA translation inhibitors and Sigma1/Sigma2 receptor regulators) showed antiviral effects, providing candidate therapeutics for repurposing.
Hunt Allcott, Luca Braghieri, Sarah Eichmeyer and Matthew Gentzkow
The authors ran a randomized experiment in which participants were paid to deactivate their Facebook accounts for four weeks before the 2018 US midterm elections. Deactivation reduced time online, increased offline socializing and TV watching, lowered factual news knowledge and political polarization, and modestly improved self-reported well-being. After the experiment, participants reduced their Facebook use, and many lowered their valuation of the platform, suggesting that habitual use exceeds its private benefits for some users.
Daniel Wrapp, Nianshuang Wang, Kizzmekia S. Corbett, Jory A. Goldsmith, Ching-Lin Hsieh, Olubukola Abiona, et al.
The authors determined a 3.5 angstrom cryo-EM structure of the SARS-CoV-2 (2019-nCoV) spike glycoprotein ectodomain stabilized in the prefusion conformation. They showed the receptor-binding domain engages human ACE2 with roughly 10- to 20-fold higher affinity than the SARS-CoV-1 spike, helping explain efficient human transmission. They also found that SARS-CoV-1 receptor-binding antibodies did not appreciably bind the new spike, informing vaccine and therapeutic design.
This paper establishes empirical scaling laws showing that the cross-entropy loss of Transformer language models follows smooth power-law relationships with model size, dataset size, and the amount of training compute. The relationships hold across many orders of magnitude, while architectural details such as width and depth have comparatively minor effects. The work provided a quantitative framework for predicting model performance and allocating compute budgets.
Google's team uses a 53-qubit programmable superconducting processor (Sycamore) to perform random quantum circuit sampling, a task chosen to be classically hard. They report sampling the output of a pseudo-random circuit in about 200 seconds, estimating that a leading classical supercomputer would need an impractically long time for the equivalent task, thereby claiming the first demonstration of quantum computational supremacy.
This paper introduces T5 (Text-to-Text Transfer Transformer), a framework that casts every NLP problem—translation, classification, question answering, summarization—as a text-to-text task with a unified model, objective, and decoding procedure. The authors conduct a large-scale empirical study comparing pre-training objectives, architectures, datasets, and transfer strategies, and release the C4 corpus. Scaling the model up to 11 billion parameters achieved state-of-the-art results on many benchmarks.
Andrew V. Anzalone, Peyton B. Randolph, Jessie R. Davis and David R. Liu
This paper introduced prime editing, a versatile genome-editing method that uses a Cas9 nickase fused to a reverse transcriptase guided by a prime editing guide RNA (pegRNA) to write new genetic information directly into a target site. Without requiring double-strand breaks or donor DNA, prime editing can install targeted insertions, deletions, and all 12 types of point mutations. The authors demonstrated correction of disease-relevant mutations in human cells with broad targeting flexibility and relatively low off-target activity.
Doruk Cengiz, Arindrajit Dube, Attila Lindner and Ben Zipperer
The authors estimate the employment effects of US state-level minimum wage increases by examining changes in the entire frequency distribution of hourly wages around each policy change. Using a bunching estimator across 138 prominent state-level increases, they count the number of jobs lost below the new minimum and the number gained at or above it. They find the number of jobs paying below the new minimum fell while jobs at or above it rose by roughly the same amount, implying minimal disemployment effects.
This paper presents the Seurat v3 framework for integrating single-cell datasets across different technologies, conditions, and modalities. It introduces 'anchors' — pairs of cells in a shared low-dimensional space representing a common biological state — to harmonize datasets and transfer labels. The methods enable joint analysis of scRNA-seq with other measurements such as protein (CITE-seq), chromatin accessibility, and spatial data.
Jacob Devlin, Ming-Wei Chang, Kenton Lee and Kristina Toutanova
BERT is a language representation model pre-trained on large unlabeled corpora using masked language modeling and next-sentence prediction, yielding deeply bidirectional contextual representations. The pre-trained model can be fine-tuned with a single additional output layer to achieve strong performance across diverse downstream tasks. It set new state-of-the-art results on eleven NLP benchmarks at the time of publication.
A. P. Drozdov, P. P. Kong, V. S. Minkov and M. I. Eremets
The authors synthesized lanthanum superhydride (LaH10) at pressures around 170 GPa and measured a superconducting transition at temperatures up to about 250 K. Evidence including the magnitude of Tc reduction in a magnetic field and an isotope effect supported phonon-mediated, conventional superconductivity. This represented a new record approaching room-temperature superconductivity in compressed hydrides.
This is the first paper in the Event Horizon Telescope (EHT) series presenting the first image of a supermassive black hole, located at the center of the galaxy M87. Using global very long baseline interferometry at 1.3 mm, the collaboration resolved a bright ring of emission surrounding a central dark region, the black hole shadow. The observations are consistent with a Kerr black hole of mass about 6.5 billion solar masses.
Alec Radford, Jeffrey Wu, Rewon Child, David Luan, Dario Amodei and Ilya Sutskever
This paper introduces GPT-2, a 1.5-billion-parameter Transformer language model trained on a large web-text corpus (WebText) with a simple next-token prediction objective. It demonstrates that a sufficiently large language model can perform many NLP tasks in a zero-shot setting, without task-specific training data or fine-tuning. The work argued that unsupervised language modeling at scale implicitly learns to perform downstream tasks from naturally occurring demonstrations.
Jae Song, David J. Price, Fatih Guvenen, Nicholas Bloom and Till von Wachter
Using US Social Security earnings records linked to employers, the authors decompose the rise in earnings inequality over 1978-2013 into within-firm and between-firm components. They find that most of the increase in earnings dispersion occurred between firms rather than within them, reflecting growing differences in average pay across employers and increased sorting of high-paid workers into high-paying firms. Within the largest firms, however, pay dispersion among workers remained relatively stable.
Armin Falk, Anke Becker, Thomas Dohmen, Benjamin Enke, David Huffman and Uwe Sunde
Using the Global Preference Survey, a dataset of risk, time, social, and trust preferences measured for about 80,000 individuals across 76 countries, the authors document how fundamental economic preferences vary across populations. They show systematic relationships between these preferences and individual behaviors as well as aggregate national outcomes. The paper also explores how geographic, cultural, and historical factors correlate with the global distribution of preferences.
Clare Bycroft, Colin Freeman and Desislava Petkova
This paper describes the open-access UK Biobank resource of deep genetic and phenotypic data on roughly 500,000 participants. It details the genotyping of ~805,000 markers, imputation to over 90 million variants, and analyses of population structure, relatedness, and genotype quality. The resource has become a foundational dataset for genome-wide association studies and human complex-trait genetics.
Nicholas Schaum, Stephen R. Quake and Tony Wyss-Coray
The Tabula Muris Consortium generated a compendium of single-cell transcriptome data spanning 20 organs and tissues from the mouse Mus musculus. Both droplet-based (10x) and FACS-sorted plate-based (Smart-seq2) methods were used to profile cells, enabling characterization of cell types across the organism. The resource provides a cross-tissue reference for defining and comparing mouse cell populations.
Raj Chetty, John N. Friedman, Nathaniel Hendren, Maggie R. Jones and Sonya R. Porter
The authors link anonymized federal tax records to census data to estimate children's adult outcomes by the neighborhood in which they grew up, at the Census-tract level for the entire United States. They build the Opportunity Atlas, showing that mobility varies enormously even between adjacent tracts and that childhood environment causally shapes later outcomes. The data reveal which specific neighborhoods foster upward mobility and identify correlates such as poverty rates, family structure, and racial composition.
The CMS Collaboration at the CERN LHC reported the observation of the Standard Model Higgs boson decaying to a pair of bottom quarks, the dominant predicted decay channel. The analysis focused on Higgs bosons produced in association with a W or Z boson (VH) in 13 TeV proton-proton collision data from 2017. Combined with earlier 7, 8, and 13 TeV measurements, the result yielded an excess at 125 GeV exceeding 5 standard deviations, establishing the decay.
Following the detection of a high-energy muon neutrino (IceCube-170922A) by the IceCube Neutrino Observatory, follow-up observations across the electromagnetic spectrum identified a spatially coincident gamma-ray flare from the blazar TXS 0506+056. The joint detection provided the first compelling evidence associating a high-energy astrophysical neutrino with a specific source, identifying blazars as cosmic-ray accelerators. The work is a landmark in multimessenger astronomy.
Victor Chernozhukov, Mert Demirer, Esther Duflo and Iván Fernández-Val
The paper develops a generic method to use any machine learning algorithm to draw valid statistical inference about features of heterogeneous treatment effects in randomized experiments. Rather than estimating the conditional average treatment effect function itself (which ML may estimate inconsistently), it targets summary parameters such as the best linear predictor of the effect on ML proxies, sorted average effects across groups, and average characteristics of the most/least affected units, using sample splitting and aggregation over many splits to obtain robust confidence intervals. It illustrates the approach with an immunization study in India.
The authors combine tax, survey, and national accounts data to build distributional national accounts that allocate 100% of U.S. national income to individuals from 1913 onward. This lets them measure income growth consistently with macroeconomic aggregates across the entire distribution, both before and after taxes and transfers. They document a sharp rise in top income shares and stagnation at the bottom since 1980.
Yuan Cao, Valla Fatemi, Shiang Fang, Kenji Watanabe, Takashi Taniguchi, Efthimios Kaxiras, et al.
The authors stacked two graphene sheets twisted by a 'magic' angle of about 1.1 degrees, producing flat electronic bands in the resulting moire superlattice. Upon electrostatically doping near half-filling of these flat bands, they observed superconductivity with critical temperatures up to about 1.7 K. The behaviour resembles that of unconventional, strongly correlated superconductors such as the cuprates, demonstrating a tunable platform for studying correlated electron physics.
Yuan Cao, Valla Fatemi, Ahmet Demir, Shiang Fang, Spencer L. Tomarken, Jason Y. Luo, et al.
This companion paper reports that twisted bilayer graphene at the magic angle exhibits insulating states at half-filling of the flat moire bands, where simple band theory would predict a metal. The authors interpret these as Mott-like correlated insulators arising from strong electron-electron interactions in the nearly flat bands. The result demonstrates that twist-angle engineering can drive graphene into strongly correlated electronic phases.
Greg Kaplan, Benjamin Moll and Giovanni L. Violante
The paper builds a Heterogeneous Agent New Keynesian (HANK) model in which households hold liquid and illiquid assets, generating realistic distributions of marginal propensities to consume. It uses this framework to analyze the transmission of monetary policy, decomposing the response of consumption into direct (intertemporal substitution) and indirect (general-equilibrium income) channels. The authors find that, unlike in representative-agent models, indirect effects operating through labor income dominate the transmission of interest rate changes.
Shannon L. Maude, Theodore W. Laetsch and Jochen Buechner
This paper reports the pivotal global phase 2 ELIANA trial of tisagenlecleucel, an anti-CD19 chimeric antigen receptor (CAR) T-cell therapy, in children and young adults with relapsed or refractory B-cell acute lymphoblastic leukemia. It demonstrated high rates of durable remission and supported the first FDA approval of a CAR-T cell therapy. The study also characterized the safety profile, including cytokine release syndrome.
Victor Chernozhukov, Denis Chetverikov, Mert Demirer, Esther Duflo, Christian Hansen, Whitney Newey, et al.
The paper develops a general framework for estimating low-dimensional treatment or structural parameters when high-dimensional nuisance components are estimated with machine learning methods. By combining Neyman-orthogonal (debiased) moment conditions with sample-splitting/cross-fitting, the approach removes regularization and overfitting biases. The resulting estimators are root-N consistent, asymptotically normal, and valid for inference despite slowly converging nuisance estimates.
Hannes Bernien, Sylvain Schwartz, Alexander Keesling, Harry Levine, Ahmed Omran, Hannes Pichler, et al.
The authors built a programmable quantum simulator using up to 51 individually trapped neutral atoms coupled to Rydberg states via optical tweezers. By tuning interactions they probed quantum many-body dynamics in a regime inaccessible to classical computation, observing the emergence of ordered antiferromagnetic phases and unexpectedly persistent, slowly relaxing oscillations after a quench. These long-lived coherent revivals later became understood as a signature of quantum many-body scars.
Nicole M. Gaudelli, Alexis C. Komor, Holly A. Rees and David R. Liu
This work developed adenine base editors (ABEs) by evolving a transfer RNA adenosine deaminase to act on DNA, enabling direct conversion of A-T base pairs to G-C in genomic DNA without double-strand breaks. Because no natural DNA adenosine deaminase was available, the authors used directed evolution to create the enzyme, then fused it to Cas9 nickase. ABEs corrected target adenines efficiently and with high product purity and low indel formation in human cells.
David Silver, Julian Schrittwieser, Karen Simonyan and Demis Hassabis
This paper presented AlphaGo Zero, which learned to play Go solely through self-play reinforcement learning without any human game data or handcrafted features, using a single neural network and a simpler tree search. Starting from random play, it discovered Go knowledge and novel strategies on its own. AlphaGo Zero surpassed all previous versions of AlphaGo, including the one that beat Lee Sedol.
This paper reports the first detection of gravitational waves from the inspiral of a binary neutron star system, event GW170817, observed on 17 August 2017 by Advanced LIGO and Advanced Virgo. The signal lasted far longer than previous black-hole mergers, consistent with low-mass compact objects. The detection, coincident with a short gamma-ray burst and subsequent electromagnetic follow-up, inaugurated multimessenger astronomy with gravitational waves.
Omar O. Abudayyeh, Jonathan S. Gootenberg, Patrick Essletzbichler and Feng Zhang
This study characterized the class 2 type VI CRISPR effector Cas13a (formerly C2c2) as a programmable RNA-targeting tool in mammalian and plant cells. A catalytically inactive Cas13a (dCas13a) was used for RNA binding while active Cas13a enabled efficient, specific knockdown of endogenous transcripts. The authors showed RNA knockdown comparable to or more specific than RNA interference and demonstrated applications such as transcript tracking and splicing modulation.
Barry Bradlyn, L. Elcoro, Jennifer Cano, M. G. Vergniory, Zhijun Wang, C. Felser, et al.
The authors develop a complete framework linking the symmetry properties of electronic bands to topological character, merging group theory of crystallographic space groups with band structure analysis. By cataloguing how atomic orbitals at Wannier centers transform under crystal symmetries, they identify which band structures are topologically trivial (atomic limit) versus topologically nontrivial. This enabled systematic, predictive identification of topological materials from symmetry data alone.
Wladimir A. Benalcazar, B. Andrei Bernevig and Taylor L. Hughes
The paper generalizes the theory of electric polarization to higher electric multipole moments (quadrupole and octupole), showing these can be topologically quantized bulk observables in crystalline insulators. It introduces tight-binding models whose gapped edges are themselves lower-dimensional topological phases, producing protected, fractionally charged corner states. This work effectively launched the field of higher-order topological insulators.
Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N. Gomez, et al.
The paper introduced the Transformer, a sequence-transduction architecture based entirely on attention mechanisms, dispensing with the recurrence and convolutions used by prior state-of-the-art models. By relying on multi-head self-attention, the model is more parallelizable and trains substantially faster, while achieving new state-of-the-art results on machine translation. The architecture became the foundation for subsequent large language models and much of modern deep learning.
Raj Chetty, David Grusky, Maximilian Hell, Nathaniel Hendren, Robert Manduca and Jimmy Narang
The authors measure absolute income mobility, defined as the fraction of children who earn more than their parents did at the same age, for U.S. birth cohorts from 1940 to the 1980s. Combining tax records with historical data, they show that this measure has declined dramatically over the period. They attribute most of the decline to the rising concentration of income growth rather than to slower aggregate growth.
Jonathan S. Gootenberg, Omar O. Abudayyeh, Jeong Wook Lee and Feng Zhang
This paper introduced SHERLOCK, a CRISPR-based nucleic acid detection platform built on the collateral RNase activity of Cas13a (C2c2). Upon recognizing a target sequence, Cas13a indiscriminately cleaves nearby reporter RNAs; combined with isothermal amplification, this yields highly sensitive, specific detection. The authors demonstrated attomolar sensitivity and single-base discrimination, with applications in detecting viruses (Zika, dengue), bacteria, and human genotypes.
The paper documents rising midlife mortality among non-Hispanic white Americans without a college degree since the late 1990s, driven by 'deaths of despair' from drugs, alcohol, and suicide alongside stalled progress against heart disease. The authors link these trends to a long-term deterioration in economic and social conditions for less-educated workers.
J. Zhang, P. W. Hess, A. Kyprianidis and C. Monroe
The authors experimentally realize a discrete (Floquet) time crystal using a chain of trapped atomic ions driven periodically into a many-body-localized regime. The system spontaneously breaks the discrete time-translation symmetry of the drive, exhibiting a subharmonic oscillation at a period that is a rational multiple of the drive period. This rigid, period-doubled response persists despite perturbations to the drive, confirming the predicted time-crystalline phase.
Salman F. Banani, Hyun O. Lee, Anthony A. Hyman and Michael K. Rosen
This review synthesizes how cells organize biochemistry into membraneless compartments termed biomolecular condensates, which form largely through liquid-liquid phase separation driven by multivalent macromolecular interactions. It describes the physical principles of condensate formation, their compositions, and how they concentrate or sequester molecules to regulate cellular processes. The authors discuss functional roles and the emerging links between aberrant condensate behavior and disease.
Nordhaus presents the updated DICE-2016R integrated assessment model and uses it to re-estimate the social cost of carbon, incorporating revised data on output, emissions, carbon cycle, and climate dynamics. The updated model yields a substantially higher social cost of carbon than earlier DICE versions and indicates that limiting warming to 2.5°C is feasible only with very rapid and stringent emissions reductions. The paper concludes that current policies fall well short of an economically optimal climate trajectory.
Grace X. Y. Zheng, Jessica M. Terry, Phillip Belgrader and Jason H. Bielas
This paper introduced a droplet-based microfluidic platform (the 10x Genomics GemCode/Chromium system) for high-throughput single-cell RNA sequencing using barcoded gel beads. The method enables 3' digital expression profiling of thousands of cells per run at low cost. The authors profiled tens of thousands of cells, including ~68,000 PBMCs, demonstrating the ability to resolve immune cell subpopulations and detect rare cell types.
Daron Acemoglu, Asuman Ozdaglar and Alireza Tahbaz-Salehi
The paper studies how the structure of production networks shapes the distribution of aggregate output, with particular attention to large downturns. It shows that even when idiosyncratic shocks are thin-tailed, interconnections can generate fat-tailed, asymmetric aggregate fluctuations in which large recessions are far more likely than large booms. Network features such as dominant suppliers and limited input substitutability amplify the probability of deep contractions.
The authors review the empirical literature on how U.S. local labor markets adjusted to the surge in import competition from China beginning around 1990. They synthesize evidence showing that the gains from trade coexisted with large, geographically concentrated adjustment costs. They conclude that markets adjusted far more slowly than standard trade models predicted.
Monkol Lek, Konrad J. Karczewski and Daniel G. MacArthur
The Exome Aggregation Consortium (ExAC) aggregates and jointly analyzes high-quality exome sequencing data from 60,706 individuals of diverse ancestries, producing the largest catalogue of human protein-coding variation at the time. The dataset reveals roughly one variant per eight exonic bases and provides direct evidence of widespread mutational recurrence. It enables improved estimation of gene-level intolerance to loss-of-function variation and refines the interpretation of pathogenic variants in clinical genetics.
Patrik L. Ståhl, Fredrik Salmén, Joakim Lundeberg and Jonas Frisén
The authors introduce 'spatial transcriptomics,' a method that places thin histological tissue sections onto a glass surface arrayed with barcoded reverse-transcription primers, so that mRNA captured at each position retains its two-dimensional spatial coordinates. Sequencing the barcoded cDNA reconstructs genome-wide expression maps directly on the tissue image. They demonstrate the approach on mouse brain and human breast cancer sections, recovering spatially resolved transcriptomes that align with tissue morphology.
Kaiming He, Xiangyu Zhang, Shaoqing Ren and Jian Sun
The authors introduced a residual learning framework that reformulates network layers to learn residual functions with reference to their inputs (via identity 'shortcut' connections), making very deep networks substantially easier to optimize. They showed that such residual networks gain accuracy from greatly increased depth, evaluating models up to 152 layers deep on ImageNet at lower complexity than VGG networks. The approach won first place in the ILSVRC 2015 classification task and yielded large improvements on detection and localization benchmarks.
Alexis C. Komor, Yongjoo B. Kim, Michael S. Packer, John A. Zuris and David R. Liu
This paper introduced cytosine base editing (CBE), a strategy that fuses a catalytically impaired Cas9 to a cytidine deaminase to directly convert C-G base pairs to T-A in genomic DNA without inducing double-strand breaks or requiring a donor template. The authors showed that within a programmable target window the deaminase converts cytosine to uracil, which is then read as thymine, and that inhibiting base-excision repair markedly improves editing efficiency. The approach achieved precise single-base correction in human and other mammalian cells.
This paper reported the first direct detection of gravitational waves, recorded on 14 September 2015 by the two LIGO interferometers as the signal GW150914. The observed waveform swept upward in frequency from 35 to 250 Hz and matched general-relativity predictions for the inspiral, merger, and ringdown of a binary black hole system, detected with a matched-filter signal-to-noise ratio of 24 and a significance greater than 5.1σ. The source was inferred to be the merger of black holes of about 36 and 29 solar masses into a final black hole of about 62 solar masses, radiating roughly 3 solar masses of energy as gravitational waves.
David Silver, Aja Huang, Chris J. Maddison and Demis Hassabis
This paper introduced AlphaGo, a system combining deep convolutional neural networks (policy and value networks) trained by supervised learning from human games and reinforcement learning by self-play, integrated with Monte Carlo tree search. The networks reduce the breadth and depth of the search needed to evaluate Go positions. AlphaGo defeated other Go programs and became the first program to beat a professional human Go player (Fan Hui) on a full-size board.
Marshall Burke, Solomon M. Hsiang and Edward Miguel
Using data from 166 countries over 1960-2010, the authors estimate a non-linear relationship between annual average temperature and economic productivity, finding output peaks near 13°C and falls sharply at higher temperatures. They project that unmitigated warming could substantially reduce average global incomes and widen global inequality by the end of the century.
The paper introduces U-Net, an encoder-decoder convolutional network with a contracting path to capture context and a symmetric expanding path with skip connections for precise localization. Combined with heavy data augmentation, the architecture trains end-to-end from very few annotated images. It won the ISBI cell-tracking and neuronal-structure segmentation challenges and segments a 512x512 image in under a second on a GPU.
Bernd Zetsche, Jonathan S. Gootenberg and Feng Zhang
This paper characterizes Cpf1 (later named Cas12a) as a single-RNA-guided DNA endonuclease of a class 2 CRISPR-Cas system, expanding the genome-editing toolbox beyond Cas9. Cpf1 requires only a single crRNA (no tracrRNA), recognizes a T-rich PAM, and produces staggered cuts with overhangs. The authors demonstrate Cpf1-mediated genome editing in human cells.
A. P. Drozdov, M. I. Eremets, I. A. Troyan, V. Ksenofontov and S. I. Shylin
The authors compressed hydrogen sulfide (H2S) to extreme pressures around 150 GPa, forming a hydrogen-rich phase believed to be H3S, and observed a sharp drop in resistance signalling superconductivity at up to 203 K. A pronounced isotope effect upon deuterium substitution indicated phonon-mediated, BCS-type (conventional) superconductivity. This set a record transition temperature at the time and validated theoretical predictions of high-Tc superconductivity in compressed hydrides.
The LHCb collaboration at CERN's Large Hadron Collider reported the observation of exotic structures in the J/ψp invariant-mass spectrum of Λ_b^0 baryon decays, interpreted as hidden-charm pentaquark states. Two resonant states, labelled Pc(4380)+ and Pc(4450)+, were identified with high statistical significance. The result provided the first compelling evidence for pentaquarks, baryon-like states composed of five quarks.
This paper introduced batch normalization, a technique that normalizes layer inputs using mini-batch statistics to reduce internal covariate shift during training. It allows higher learning rates and less careful initialization, accelerates convergence, and acts as a regularizer. Applied to image classification networks, it dramatically reduced training steps and improved accuracy.
This paper introduced Adam, a first-order gradient-based optimization algorithm for stochastic objective functions that computes adaptive per-parameter learning rates from estimates of the first and second moments of the gradients. The method is computationally efficient, has low memory requirements, and is well suited to large-scale and noisy/sparse-gradient problems. It became one of the most widely used optimizers in deep learning.
Kok Hao Chen, Alistair N. Boettiger, Jeffrey R. Moffitt, Siyuan Wang and Xiaowei Zhuang
This paper introduces MERFISH, a single-molecule imaging method that uses error-robust barcoding combined with sequential rounds of hybridization and imaging to identify many RNA species in single cells in situ. The error-correcting binary codes allow detection and misidentification correction across thousands of transcripts. The authors imaged up to ~1,000 genes in individual human cells, mapping spatial expression and revealing gene co-variation and subcellular RNA localization.
Volodymyr Mnih, Koray Kavukcuoglu and David Silver
The paper introduced the Deep Q-Network (DQN), which combines Q-learning with deep convolutional networks and stabilizing techniques such as experience replay and a target network. Trained end-to-end from raw pixels and game scores, a single architecture and hyperparameter set learned to play 49 Atari 2600 games. It reached or exceeded the level of a professional human games tester on the majority of titles, demonstrating a general agent learning directly from high-dimensional sensory input.
The ATLAS Collaboration reported the observation of a new neutral boson in searches for the Standard Model Higgs boson, using proton-proton collision data from the LHC corresponding to about 4.8 fb⁻¹ at √s = 7 TeV (2011) and 5.8 fb⁻¹ at √s = 8 TeV (2012). Combining sensitive decay channels, the analysis found an excess at a mass of about 126.0 GeV with a local significance of 5.9 standard deviations, consistent with the production and decay of a Standard Model Higgs boson. The result, announced jointly with a companion CMS measurement, provided experimental confirmation of the boson associated with the Brout–Englert–Higgs mechanism of electroweak symmetry breaking.
Martin Jinek, Krzysztof Chylinski, Ines Fonfara, Michael Hauer, Jennifer A. Doudna and Emmanuelle Charpentier
This study demonstrated that the CRISPR-associated protein Cas9 from Streptococcus pyogenes is an RNA-guided DNA endonuclease whose target specificity is determined by a dual-RNA structure formed by a CRISPR RNA (crRNA) base-paired to a trans-activating crRNA (tracrRNA). The authors showed that Cas9 introduces site-specific double-strand breaks in target DNA, with its HNH domain cleaving the complementary strand and its RuvC-like domain cleaving the noncomplementary strand. Critically, they engineered the two guide RNAs into a single chimeric guide RNA that still directed sequence-specific cleavage, establishing the system as a programmable tool for genome editing.
Takahashi and Yamanaka demonstrated that pluripotent stem cells can be generated directly from mouse fibroblast cultures by introducing a defined set of transcription factors. Screening candidate genes associated with pluripotency, they identified four factors—Oct3/4, Sox2, c-Myc, and Klf4—that were sufficient to reprogram both embryonic and adult fibroblasts into cells they termed induced pluripotent stem (iPS) cells. These iPS cells resembled embryonic stem cells in morphology, growth, and marker expression and could form teratomas containing tissues of all three germ layers.
This paper demonstrated that atomically thin, monocrystalline graphitic films—down to a single atomic layer (graphene)—can be isolated, are stable under ambient conditions, and are of high crystalline and electronic quality. The authors showed that these films behave as a two-dimensional semimetal with a small overlap between valence and conduction bands and exhibit a strong ambipolar electric field effect, with charge-carrier concentrations up to 10¹³ cm⁻² and room-temperature mobilities of about 10,000 cm²/V·s tunable by a gate voltage. The work launched the experimental field of two-dimensional materials and underpinned the 2010 Nobel Prize in Physics awarded to Geim and Novoselov.
Eric S. Lander, Lauren M. Linton, Bruce Birren, Chad Nusbaum and Michael C. Zody
This paper reported the results of the publicly funded Human Genome Project, presenting and making freely available a draft sequence covering the great majority of the human genome along with an initial analysis. The consortium described the broad genomic landscape—including gene content, repeat elements, GC content, and recombination rates—and estimated a surprisingly low number of protein-coding genes, on the order of roughly 30,000–40,000. The work provided a foundational reference for human biology, medicine, and evolutionary studies.
Randall K. Saiki, Stephen Scharf, Fred Faloona, Kary B. Mullis, Glenn T. Horn, Henry A. Erlich, et al.
This paper reported the first published application of in vitro primer-mediated enzymatic amplification of DNA—the technique that became known as the polymerase chain reaction (PCR)—as part of a rapid, sensitive prenatal diagnostic test for sickle cell anemia. The authors amplified specific β-globin target sequences from genomic DNA roughly 220,000-fold and then distinguished the normal (βA) and sickle (βS) alleles by restriction endonuclease digestion of a hybridized end-labeled oligonucleotide probe. The combined procedure allowed genotyping in under a day using far less than one microgram of genomic DNA.
Françoise Barré-Sinoussi, Jean-Claude Chermann, F. Rey, M. T. Nugeyre, S. Chamaret and Luc Montagnier
The authors reported the isolation of a novel retrovirus from a lymph-node biopsy of a patient with lymphadenopathy considered at risk for AIDS. The virus exhibited magnesium-dependent reverse transcriptase activity, budded as a type-C particle, and showed tropism for T lymphocytes, but was antigenically and biologically distinct from the known human T-cell leukemia viruses (HTLV-I/II). This agent—later designated lymphadenopathy-associated virus (LAV) and subsequently recognized as HIV—was proposed as a candidate cause of AIDS.
Kahneman and Tversky presented experimental evidence that people systematically violate the axioms of expected utility theory when choosing among risky prospects, and proposed prospect theory as a descriptive alternative. In their model, outcomes are evaluated as gains and losses relative to a reference point through an S-shaped value function that is concave for gains, convex for losses, and steeper for losses (loss aversion), while objective probabilities are transformed by a nonlinear decision-weighting function that overweights small probabilities. The theory accounts for observed anomalies such as the certainty, reflection, and isolation effects.
Sanger, Nicklen, and Coulson introduced a new method for determining nucleotide sequences in DNA, building on their earlier 'plus and minus' technique. The method uses 2′,3′-dideoxy and arabinonucleoside analogues of the normal deoxynucleoside triphosphates, which act as specific chain-terminating inhibitors of DNA polymerase, generating a set of partially extended chains that can be size-separated by gel electrophoresis to read the sequence. Applied to bacteriophage φX174 DNA, the approach proved faster and more accurate than the original plus or minus method.
Using extracellular single-unit recordings from the hippocampus of freely moving rats, O'Keefe and Dostrovsky observed that certain neurons fired selectively when the animal occupied particular locations or orientations within the environment. They interpreted these spatially tuned responses as preliminary evidence that the hippocampus functions as a spatial map of the animal's surroundings. This brief report introduced the concept of hippocampal 'place cells' and seeded the cognitive-map theory of hippocampal function.
Manabe and Wetherald developed a one-dimensional radiative-convective model of the atmosphere that, given a fixed distribution of relative humidity, computed the vertical temperature profile in thermal equilibrium. Incorporating realistic radiative transfer for water vapor, carbon dioxide, and ozone together with a convective adjustment, they estimated the surface and atmospheric temperature response to changes in CO₂. The study is widely regarded as the first physically based estimate of equilibrium climate sensitivity and demonstrated the central role of water-vapor feedback in amplifying CO₂-induced warming.
Penzias and Wilson reported that the 20-foot horn-reflector antenna at Bell Telephone Laboratories in Holmdel, New Jersey, measured an excess zenith antenna temperature of about 3.5 K at 4080 Mc/s (4.08 GHz) that could not be attributed to known sources of noise. They found this excess radiation to be isotropic, unpolarized, and free of seasonal variation within their measurement limits. In a companion paper, Dicke and colleagues interpreted this signal as relic radiation from a hot early universe, and it is now recognized as the discovery of the cosmic microwave background, providing key observational support for the Big Bang model.
In this one-page report, Watson and Crick proposed a double-helical structure for the salt of deoxyribose nucleic acid (DNA), consisting of two right-handed helical polynucleotide chains coiled around a common axis and running in antiparallel directions. They proposed that the chains are held together by hydrogen bonding between specific complementary base pairs—adenine with thymine and guanine with cytosine—a feature dictated by the structure. They famously noted that this specific pairing immediately suggested a possible copying mechanism for the genetic material.
Drawing on voltage-clamp measurements of ionic currents in the squid giant axon, Hodgkin and Huxley developed a quantitative mathematical model describing membrane current as the sum of separate sodium, potassium, and leak conductances that vary with voltage and time. Using a system of nonlinear differential equations with voltage-dependent gating variables, they reproduced the form, amplitude, and conduction velocity of the action potential and other excitation phenomena. The model unified their experimental findings and became the foundational framework for quantitative electrophysiology.
In this hospital-based case-control study, Doll and Hill interviewed roughly 700 patients with carcinoma of the lung and a comparison group of patients with other diseases across London hospitals to assess the association between tobacco smoking and lung cancer. They found that lung-cancer patients were substantially more likely to be smokers and to be heavy smokers than the controls, with risk rising with the amount smoked. The authors concluded that smoking is an important factor in the production of carcinoma of the lung.
Shannon established the mathematical foundations of information theory, modeling communication as the reproduction of a message selected from a set of possibilities across a (possibly noisy) channel. He introduced entropy as a quantitative measure of information and uncertainty, defined channel capacity, and proved coding theorems showing that reliable transmission is achievable up to (but not beyond) the channel capacity. The work unified the treatment of discrete and continuous sources and laid the groundwork for modern digital communication and data compression.
Turing introduced an abstract model of computation—later called the Turing machine—to make precise the notion of a function being effectively calculable, defining 'computable numbers' as those whose decimals can be produced by such a machine. He proved the existence of a universal machine able to simulate any other and showed that some well-defined problems, including the halting problem, are not computable. Using these results he demonstrated that Hilbert's Entscheidungsproblem (decision problem) for first-order logic has no general algorithmic solution.
Fleming reported that a mould of the genus Penicillium, which had contaminated a culture plate, produced a diffusible substance—which he named 'penicillin'—that inhibited the growth of many common pathogenic bacteria. He characterized the antibacterial spectrum and potency of the crude culture filtrate, showing it was strongly active against Gram-positive organisms (such as staphylococci and streptococci) while sparing certain Gram-negative bacilli. He proposed using the agent both as a selective culture medium for isolating Bacillus influenzae and as a relatively non-toxic antiseptic.
Hubble combined distances to roughly two dozen extra-galactic nebulae (galaxies), estimated largely from Cepheid variables and other stellar indicators, with their measured radial velocities to test for a systematic relationship. He found an approximately linear correlation in which the radial velocity of a nebula increases with its distance, with a proportionality constant of about 500 km/s per megaparsec. This velocity–distance relation, now known as Hubble's law, provided the first observational evidence that the universe is expanding and became a cornerstone of modern cosmology.
In this foundational paper ("On the Electrodynamics of Moving Bodies") Einstein introduced the special theory of relativity, building physics on two postulates: that the laws of physics are identical in all inertial frames and that the speed of light in vacuum is the same for all observers regardless of the motion of the source. From these he derived the Lorentz transformation kinematically, abolishing the need for a luminiferous ether and showing that simultaneity, lengths, and time intervals are relative to the observer's frame of reference. The work reconciled mechanics with Maxwell's electrodynamics and established time dilation and length contraction as physical consequences of the structure of spacetime.