Papersby GlyphLatest papers
The brief · Mon Nov 9The unverified paper "VFlash: Fast, Accurate Visual Planning via Flash Primitive Decoding" by NVIDIA speeds up visual motion planning for robots by predicting spatial shapes at once, offering leaders lower robotics deployment latency. Carnegie Mellon University released the unverified paper "Random Is Almost Enough? Stable and Low-Cost Benchmark Compression via BestRand", showing small benchmark subsets reliably estimate full scores to lower evaluation compute costs. ServiceNow introduced the unverified paper "Apriel Voice 27B Preview: An Open Model for Speech, Reasoning, and Tool Use", demonstrating an open model that directly processes spoken speech to converse, reason, and execute tool calls for voice workloads.
Read more
Thursday, Nov 51 paper
VFlash: Fast, Accurate Visual Planning via Flash Primitive DecodingA method called VFlash speeds up and improves visual motion planning for robots by predicting entire spatial shapes at once.Monday, Nov 23 papers
MoviEdit: Long-form Multi-Shot Video Editing Made EasyMoviEdit is a framework and benchmark for editing long videos across multiple camera shots and changing scenes.Saturday, Oct 311 paper
Random Is Almost Enough? Stable and Low-Cost Benchmark Compression via BestRandBestRand is a method for selecting small benchmark subsets that reliably estimate full evaluation scores at lower computational costs.Thursday, Oct 291 paper
Apriel Voice 27B Preview: An Open Model for Speech, Reasoning, and Tool UseApriel Voice 27B is an open AI model that directly processes spoken speech to converse, reason, and use external tools.Tuesday, Oct 273 papers
ThinkWorld: World Imagination Reinforces Chain-of-Dynamics ReasoningThinkWorld improves robotic generalization by representing actions as dynamics chains and using world imagination to learn from unlabeled human videos.Monday, Oct 262 papers
StreamVibe: Proactive Video Agents for Evolving User IntentStreamVibe introduces a benchmark, streaming vision-language model, and hierarchical memory harness to enable proactive, intent-aware interaction with implicit and evolving user goals.Sunday, Oct 251 paper
InteractiveWorld: A Verifiable Human-Centric Environment for Building Better Interactive MLLM AssistantsInteractiveWorld provides a verifiable simulation environment and benchmark to evaluate whether multimodal language models can reliably guide humans through physical tasks.Saturday, Oct 242 papers
Eat Your Precision: Precision Objective Mitigates Reward Hacking in Rubric-based RLFORK jointly optimizes recall and precision via rubric-driven necessity credits and factual grading to improve response quality and reduce verbosity in reinforcement learning.Friday, Oct 233 papers
On-Policy Self-Training: Data-Free Self-Improvement in Large Language ModelsFoundation models can self-improve across text and vision tasks by generating synthetic questions and reasoning traces guided only by high-level topics, without external supervision.Thursday, Oct 222 papers
UserArena: Benchmarking Interactive User Simulation for Long-Horizon Shopping AgentsUserArena introduces a real-world interactive benchmark with long-horizon shopping trajectories and an evaluation framework, alongside On-Policy Trajectory Distillation to improve simulator fidelity for agentic recommendations.Wednesday, Oct 211 paper
Retrieval Augmented Generation CompositionalityThis study shows that individual RAG techniques do not always work well together and provides a framework to evaluate full retrieval pipelines.Tuesday, Oct 202 papers
Joints, Camera, Action! Co-Generating 3D Motion and Video for Arbitrary RigsThis model generates both video and 3D motion for any character rig, allowing the motion to be replayed accurately within a 3D scene.Monday, Oct 193 papers
OpenVLHarness: Real World Visual Agentic Tool-Use with Multimodal MemoryThis framework improves multimodal AI reasoning by providing persistent memory and tool interfaces to help models process and combine complex visual evidence.Friday, Oct 161 paper
Self-Evolving Coding Agents via Harness-Policy OptimizationThis framework improves coding agents by recursively alternating between optimizing the agent's policy and the environment harness it uses to interact with code.Wednesday, Oct 144 papers
Correction Training for Diffusion ModelsThis method adds a correction step to diffusion models to fix errors that accumulate during the image generation process.Tuesday, Oct 1312 papers
AesCode: Aesthetic Code Generation with Decoupled Cross-Modal RewardsThis paper introduces a method to improve code-based visual generation by using image-based aesthetic guidance and decoupled rewards to ensure quality and correctness.Monday, Oct 121 paper
X-Ragent: Tool-Augmented Reinforcement Learning for Evidence-Grounded Chest X-Ray ReasoningThis paper introduces a reinforcement learning framework for medical agents that selectively uses external evidence to improve diagnostic reasoning in chest X-rays.Saturday, Oct 101 paper
Last Secure Code Benchmark: Is the Code You Vibe Secure?This paper introduces a benchmark showing that current coding agents frequently produce functional but insecure code, highlighting significant security risks in automated coding.Friday, Oct 93 papers
Scaling Agentic Data for Long-Horizon Terminal IntelligenceThis paper introduces a system to automatically generate high-quality training data for long-horizon terminal agents to improve their autonomous problem-solving capabilities.Thursday, Oct 82 papers
Thinking Inertia: LLMs Keep Thinking When Told Not ToThis paper shows that explicit controls often fail to stop LLMs from generating internal reasoning traces, a phenomenon the authors call thinking inertia.Wednesday, Oct 72 papers
Long-WAM: Scaling the Context of World-Action ModelsLong-WAM improves robot control by using autoregressive pretraining to effectively leverage long visual histories for real-time action prediction.Tuesday, Oct 66 papers
Self-Retrospection Distillation: Turning Post-hoc Experiences into Prior ForesightSelf-Retrospection Distillation improves agent performance by using hindsight from completed trajectories to supervise foresight predictions during training.Monday, Oct 52 papers
LoGRA: Scaling LLM Reinforcement Learning with Low-Rank Gradient SketchesLoGRA reduces memory usage during reinforcement learning by using low-rank gradient sketches and step control to stabilize training.Saturday, Oct 31 paper
ALoDLM: Adaptively Looped Diffusion Language ModelsALoDLM improves diffusion language models by using token-adaptive latent recurrence to allocate more computation to difficult tokens.
Amazon1 citations80 upvotesRead →Friday, Oct 22 papers
Native Action-Prior Learning from Videos for World Action ModelsNAVA-WAM learns robot action priors directly from observation-only videos, improving control performance without needing action-labeled data for pretraining.Thursday, Oct 18 papers
Sharpening Tax in Post-TrainingThis paper introduces a metric to quantify the loss in solution coverage after reinforcement learning and proposes a sampler to mitigate it.Wednesday, Sep 302 papers
Mid-Harness: Scaling Actions Between Model and Harness for Terminal AgentsMid-Harness improves agent reliability by sampling and verifying candidate actions before execution, rather than relying on a single generation.Tuesday, Sep 299 papers
Predictive Credit: Measuring What Scientific Explanations Add to Experimental ForecastsThis paper introduces a protocol to measure how much scientific explanations actually improve the accuracy of experimental forecasts.Amazon38 upvotesRead →PixelUMM: Encoder-Free Unified Image and Video Understanding and GenerationPixelUMM is an encoder-free model that unifies image and video understanding and generation directly in pixel space.Monday, Sep 283 papers
Rethinking Latent Visual Reasoning: Grounding Latent Reasoning in Visual EvidenceReaLVR improves latent visual reasoning by providing explicit supervision to latent trajectories, helping models focus on relevant visual evidence.Amazon2 citations210 upvotesRead →LEGO-Anything: Coding Agents for 3D Scene ReconstructionLEGO-Anything is a framework where an agent iteratively writes and executes code to reconstruct 3D scenes from images.
AWS135 upvotesRead →An RL View of OPD: Least Square Policy Distillation for Sample-Efficient LLM ReasoningLSPD applies reinforcement learning principles to policy distillation, improving sample efficiency and policy diversity in language model training.Sunday, Sep 273 papers
QwenGyre: An Elastic Reinforcement Learning Framework for Training xLong-Horizon AgentsQwenGyre is an online reinforcement learning framework that elastically manages GPU resources to train agents on extremely long-horizon tasks.Saturday, Sep 261 paper
In-Flight KV Cache with Clean Anchors for Faster Autoregressive Video DiffusionFlashForward accelerates video diffusion by reusing in-flight KV caches and using sparse clean anchors to maintain structural consistency.Wednesday, Sep 232 papers
Training Object Permanence in World ModelsThis paper introduces a dataset and training stack to improve object permanence and physical reasoning in video world models.Tuesday, Sep 223 papers
The Tasteful Agent: Measuring and Improving Taste in Long-Horizon TasksTaste-Bench measures an agent's ability to make good long-horizon decisions by evaluating choices at critical decision forks in trajectories.Monday, Sep 213 papers
RRSI: Regularized Recursive Self-Improvement of Agent HarnessesRRSI keeps self-improving agent harnesses from memorising their training tasks, so gains carry over to new benchmarks.Sunday, Sep 201 paper
One to More, More to One: Category-Aware Iterative Expert Training for Software Engineering AgentsThis framework improves software engineering agents by training category-specific experts and consolidating them into a single student model.Friday, Sep 184 papers
OmniVChat: Synthesizing, Benchmarking, and Training for Native Audio-Visual DialogueOmniVChat is a framework for native audio-visual dialogue that includes a data engine for synthesis and a reinforcement learning reward design.Thursday, Sep 176 papers
DeepSeek-V4.1-Flash: Pushing the Limits of KV Cache CompressionA 552B mixture-of-experts model with a 1M-token context, built to shrink the KV cache for agent workloads.Amazon44 upvotesRead →FAMOS: Feed-Forward 3D Articulation Modeling from Sparse ObservationsFAMOS is a feed-forward model that predicts articulated object motion from sparse point cloud observations.Wednesday, Sep 161 paper
BI-Agent and BI-Bench: Towards Automating End-to-End Business IntelligenceBI-Agent automates end-to-end business intelligence tasks by decomposing workflows into subtasks like data search, join, and transformation.Tuesday, Sep 151 paper
Verifiable Social Reasoning for LLM AssistantsFuse is a multi-agent simulation framework that allows for the verifiable evaluation of an assistant's social reasoning capabilities.Monday, Sep 141 paper
Dream-RSI: Recursive Self-Improvement through Evolving WorldsDream-RSI enables scalable recursive self-improvement by using historical discovery replay to evaluate exploration policies offline, reducing costly online evaluations.Sunday, Sep 131 paper
Flattening Every Memory Peak in Long-Context Mixture-of-Experts TrainingBounds the four memory peaks that crash long-context mixture-of-experts training, from expert dispatch to optimizer state, with fixed-size GPU schedules.Friday, Sep 112 papers
SteerDuplex: Steerable Duplex Speech Dialogue ModelsA full-duplex speech model that follows spoken instructions on tone, persona, pace and voice, with a benchmark of 390 prompts to test it.Thursday, Sep 102 papers
Benchmark Radar: A Living Database and Search Engine for AI Benchmarks and EvaluationBenchmark Radar is a searchable living database and discovery engine for AI evaluation benchmarks that aggregates sources, score histories, and evidence to support benchmark selection and comparison.AWS1 citations33 upvotesRead →Wednesday, Sep 93 papers
An Open Recipe for IMO Gold: Training Nemotron for Olympiad MathematicsA natural-language proof-generation pipeline using post-trained Nemotron 3 Ultra checkpoints achieves gold-medal performance on IMO 2026 through iterative verification and refinement without external tools.Tuesday, Sep 81 paper
Procedural Graphs: Self-Evolving Execution Structures for LLM AgentsA procedural graph framework organizes agent actions into structured relational triplets, providing situational guidance and self-evolving topology to improve long-horizon tool use.Monday, Sep 71 paper
Online Draft Co-Training for Speculative Decoding in Large-Scale, Long-Context RL Post-TrainingA system for large-scale online draft co-training accelerates speculative decoding in RL post-training by extending context-parallel attention and adding cross-stage feature transport.Most cited
Semantic ScholarAttention Is All You NeedIntroduced the Transformer, the attention-only architecture behind nearly every large language model since.
Google201k citations142 upvotesRead →BERT: Pre-training of Deep Bidirectional Transformers for Language UnderstandingPre-training a Transformer to fill in masked words, then fine-tuning it, beat task-specific models across language understanding tests.Google125k citations33 upvotesRead →Language Models are Few-Shot LearnersGPT-3, a 175B-parameter model, does new tasks from a few examples in the prompt, with no fine-tuning.
OpenAI66.4k citations21 upvotesRead →GPT-4 Technical ReportA multimodal model that passes a simulated bar exam with a score around the top 10% of test takers.OpenAI28.4k citations7 upvotesRead →Training language models to follow instructions with human feedbackInstructGPT: fine-tuning on human feedback made a 1.3B model preferred over the 175B GPT-3.OpenAI25.8k citations26 upvotesRead →LoRA: Low-Rank Adaptation of Large Language ModelsLoRA fine-tunes a large model by training small low-rank matrices, cutting trainable parameters by 10,000 times.
Microsoft25.6k citations66 upvotesRead →Chain-of-Thought Prompting Elicits Reasoning in Large Language ModelsAsking a model to write out its intermediate steps (chain of thought) sharply improves its math and logic answers.Google23.8k citations16 upvotesRead →LLaMA: Open and Efficient Foundation Language ModelsOpen models from 7B to 65B trained only on public data; the 13B model beats GPT-3 on most benchmarks.
Meta22.6k citations25 upvotesRead →The Llama 3 Herd of ModelsOpen models up to 405B parameters that match leading closed models on many tasks.Meta20.8k citations119 upvotesRead →Segment AnythingA promptable model and a dataset of over a billion masks that cut out any object in any image.Meta16.8k citations6 upvotesRead →Direct Preference Optimization: Your Language Model is Secretly a Reward ModelDPO aligns a model to human preferences with a simple classification loss, no reward model or RL loop.
Stanford University11.6k citations73 upvotesRead →DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language ModelsDeepSeekMath 7B improves mathematical reasoning through enhanced data pre-training and Group Relative Policy Optimization, achieving high scores on MATH benchmark without external tools.
DeepSeek10.6k citations152 upvotesRead →Mamba: Linear-Time Sequence Modeling with Selective State SpacesA selective state space model that scales linearly with sequence length and matches Transformers on language.
Carnegie Mellon University10.2k citations152 upvotesRead →Qwen3 Technical ReportQwen3, a unified series of large language models, integrates thinking and non-thinking modes, reduces computational resources, and achieves state-of-the-art performance across various tasks and languages.
Alibaba (Qwen)10.1k citations339 upvotesRead →Efficient Memory Management for Large Language Model Serving with PagedAttentionvLLM manages the KV cache like pages of virtual memory, serving models with 2 to 4 times the throughput.
UC Berkeley9,965 citations72 upvotesRead →BLIP-2: Bootstrapping Language-Image Pre-training with Frozen Image Encoders and Large Language ModelsBLIP-2 efficiently pre-trains vision-language models using off-the-shelf frozen encoders and decoders, achieving state-of-the-art performance with fewer parameters.
Salesforce9,805 citations3 upvotesRead →Scaling Laws for Neural Language ModelsLanguage model loss falls as a smooth power law as model size, data and compute grow.OpenAI9,592 citations12 upvotesRead →BLIP: Bootstrapping Language-Image Pre-training for Unified Vision-Language Understanding and GenerationBLIP, a Vision-Language Pre-training framework, improves performance across both understanding and generation tasks by bootstrapping captions from noisy web data, achieving state-of-the-art results on image-text retrieval, image captioning, and VQA, and showing strong generalization to video-language tasks.Salesforce7,655 citations5 upvotesRead →DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement LearningReinforcement learning taught a model long step-by-step reasoning on par with OpenAI o1.DeepSeek5,972 citations458 upvotesRead →FlashAttention: Fast and Memory-Efficient Exact Attention with IO-AwarenessFlashAttention computes exact attention with far fewer GPU memory reads and writes, making long sequences faster.Stanford University5,830 citations16 upvotesRead →GPT-4o System CardGPT-4o is an omnimodal autoregressive model trained to handle text, audio, image, and video inputs, offering high-performance outputs across these modalities, with particular strengths in vision and audio.OpenAI5,388 citations88 upvotesRead →Gemini 2.5: Pushing the Frontier with Advanced Reasoning, Multimodality, Long Context, and Next Generation Agentic CapabilitiesGemini 2.X model family, including Gemini 2.5 Pro and Flash, offers superior coding, reasoning, and multimodal understanding capabilities across a range of computational efficiencies.Google4,594 citations67 upvotesRead →Gemini 1.5: Unlocking multimodal understanding across millions of tokens of contextRecalls details across millions of tokens of text, hours of video and days of audio.
Google DeepMind4,190 citations65 upvotesRead →Training Compute-Optimal Large Language ModelsChinchilla: for a fixed compute budget, train a smaller model on more data; parameters and tokens should grow together.Google DeepMind4,035 citations13 upvotesRead →Constitutional AI: Harmlessness from AI FeedbackConstitutional AI trains a harmless assistant from AI feedback guided by a short list of written principles, not human harm labels.
Anthropic3,987 citations4 upvotesRead →Megatron-LM: Training Multi-Billion Parameter Language Models Using Model ParallelismMegatron-LM splits each Transformer layer across GPUs to train models with billions of parameters.
NVIDIA3,387 citations5 upvotesRead →Qwen3-VL Technical ReportQwen3-VL, a vision-language model, excels in text and multimodal understanding through advanced architectures and larger contexts, achieving superior performance across benchmarks.Alibaba (Qwen)3,178 citations164 upvotesRead →BLOOM: A 176B-Parameter Open-Access Multilingual Language ModelBLOOM, a 176B-parameter open-access language model, achieves competitive performance on various benchmarks and is released publicly with a Responsible AI License.
Hugging Face3,165 citations41 upvotesRead →Phi-3 Technical Report: A Highly Capable Language Model Locally on Your PhoneSmall models trained on filtered and synthetic data that run on a phone and rival much larger ones.Microsoft2,649 citations262 upvotesRead →CodeT5: Identifier-aware Unified Pre-trained Encoder-Decoder Models for Code Understanding and GenerationCodeT5, a unified encoder-decoder Transformer model, improves code understanding and generation by leveraging semantic information from identifiers and user comments.Salesforce2,633 citations3 upvotesRead →From Local to Global: A Graph RAG Approach to Query-Focused SummarizationGraphRAG builds a knowledge graph of a document set so a model can answer questions about the whole collection.Microsoft2,561 citations7 upvotesRead →VGGT: Visual Geometry Grounded TransformerVGGT, a feed-forward neural network, efficiently infers multiple 3D attributes from single or multiple views, outperforming alternatives and enhancing downstream tasks without post-processing.Meta2,132 citations40 upvotesRead →DeepSeek-Coder: When the Large Language Model Meets Programming -- The Rise of Code IntelligenceDeepSeek-Coder, a series of open-source code models trained from scratch, outperforms both open-source and closed-source models across multiple benchmarks.DeepSeek2,033 citations74 upvotesRead →DINOv3DINOv3, a self-supervised learning model, achieves superior performance across various vision tasks by scaling datasets and models, addressing dense feature degradation, and enhancing flexibility with post-hoc strategies.Meta1,932 citations320 upvotesRead →SAM 3: Segment Anything with ConceptsSegment Anything Model 3 achieves state-of-the-art performance in promptable concept segmentation and tracking by leveraging a unified model architecture with decoupled recognition and localization.Meta1,466 citations138 upvotesRead →DeepSeekMoE: Towards Ultimate Expert Specialization in Mixture-of-Experts Language ModelsThe DeepSeekMoE architecture improves expert specialization in Mixture-of-Experts models by segmenting experts and isolating shared ones, achieving better performance and computational efficiency compared to GShard and other models.DeepSeek1,268 citations65 upvotesRead →WizardCoder: Empowering Code Large Language Models with Evol-InstructWizardCoder, a Code LLM fine-tuned with complex instructions using Evol-Instruct, outperforms other open-source and closed LLMs on several code generation benchmarks.Microsoft1,086 citations34 upvotesRead →DeepSeek-VL: Towards Real-World Vision-Language UnderstandingDeepSeek-VL is an open-source vision-language model that achieves state-of-the-art performance in real-world applications by combining a hybrid vision encoder with effective pretraining strategies to preserve language model capabilities.DeepSeek939 citations49 upvotesRead →The Impact of AI on Developer Productivity: Evidence from GitHub CopilotIn a controlled experiment, developers with GitHub Copilot finished a programming task 55.8% faster than those without it.
GitHub906 citationsRead →DeepSeek-V3.2: Pushing the Frontier of Open Large Language ModelsDeepSeek-V3.2 introduces DeepSeek Sparse Attention and a scalable reinforcement learning framework, achieving superior reasoning and performance compared to GPT-5 and Gemini-3.0-Pro in complex reasoning tasks.DeepSeek887 citations275 upvotesRead →DeepSeek LLM: Scaling Open-Source Language Models with LongtermismDeepSeek LLM, an open-source language model project, develops a large dataset and employs SFT and DPO to achieve performance surpassing LLaMA-2 70B and GPT-3.5 in various benchmarks and open-ended evaluations.DeepSeek872 citations56 upvotesRead →CodeT5+: Open Code Large Language Models for Code Understanding and GenerationCodeT5+, a family of flexible encoder-decoder LLMs initialized with frozen LLMs and trained with a mixture of pretraining objectives, achieves state-of-the-art performance across various code-related tasks.Salesforce836 citations7 upvotesRead →Datasets: A Community Library for Natural Language ProcessingDatasets is a community library for NLP that standardizes interfaces, versioning, and documentation for over 650 datasets with contributions from more than 250 individuals.Hugging Face828 citations16 upvotesRead →Group Sequence Policy OptimizationGroup Sequence Policy Optimization (GSPO) is a reinforcement learning algorithm that improves training efficiency and performance of large language models by using sequence-level importance ratios and operations.Alibaba (Qwen)814 citations323 upvotesRead →Back to Basics: Revisiting REINFORCE Style Optimization for Learning from Human Feedback in LLMsReinforcement Learning from Human Feedback (RLHF) benefits from simpler REINFORCE-style optimization methods over computationally expensive PPO or "RL-free" methods, enhancing performance and reducing cost in large language models.
Cohere804 citations18 upvotesRead →Sleeper Agents: Training Deceptive LLMs that Persist Through Safety TrainingModels trained with a hidden backdoor kept their deceptive behaviour through standard safety training.Anthropic657 citations30 upvotesRead →Qwen3-Omni Technical ReportQwen3-Omni, a multimodal model, achieves state-of-the-art performance across text, image, audio, and video, using a Thinker-Talker MoE architecture and a lightweight causal ConvNet for efficient streaming synthesis.Alibaba (Qwen)653 citations155 upvotesRead →Native Sparse Attention: Hardware-Aligned and Natively Trainable Sparse AttentionNSA, a trainable sparse attention mechanism, enhances long-context modeling efficiency without sacrificing performance, achieving improvements in speed and accuracy over full attention models.DeepSeek596 citations157 upvotesRead →DeepSeek-Coder-V2: Breaking the Barrier of Closed-Source Models in Code IntelligenceDeepSeek-Coder-V2, a Mixture-of-Experts language model, excels in code-specific tasks by enhancing coding and mathematical reasoning capabilities while expanding language support and context length.DeepSeek530 citations73 upvotesRead →Janus: Decoupling Visual Encoding for Unified Multimodal Understanding and GenerationJanus, an autoregressive framework with separate visual encoding pathways within a unified transformer architecture, enhances performance in unified multimodal understanding and generation.DeepSeek525 citations37 upvotesRead →LoRA Learns Less and Forgets LessLoRA learns less than full fine-tuning on code and math, but forgets less of what the model already knew.
Databricks486 citations91 upvotesRead →WorkArena: How Capable Are Web Agents at Solving Common Knowledge Work Tasks?WorkArena and BrowserGym evaluate large language model-based agents' ability to perform enterprise software tasks, revealing gaps in current agent capabilities and differences between open and closed-source LLMs.
ServiceNow443 citations2 upvotesRead →Aya Model: An Instruction Finetuned Open-Access Multilingual Language ModelAya, a multilingual generative language model supporting over 50% lower-resourced languages, excels in both generative and discriminative tasks across 99 languages, and provides extensive evaluations and open-source resources.Cohere438 citations48 upvotesRead →Rubrics as Rewards: Reinforcement Learning Beyond Verifiable DomainsRubrics as Rewards (RaR) framework uses structured rubrics as interpretable reward signals for on-policy training, improving performance in real-world reinforcement learning tasks with subjective criteria.
Scale AI431 citations6 upvotesRead →Replacing Judges with Juries: Evaluating LLM Generations with a Panel of Diverse ModelsEvaluating Large Language Models using a Panel of smaller evaluators outperforms single large evaluators with reduced bias and cost.Cohere418 citations71 upvotesRead →Alignment faking in large language modelsClaude 3 Opus sometimes went along with a training goal it disagreed with, to avoid being changed, without being told to.Anthropic393 citations10 upvotesRead →Why Language Models HallucinateModels hallucinate because training and benchmarks reward confident guessing over saying they do not know.OpenAI391 citations200 upvotesRead →SWE-Bench Pro: Can AI Agents Solve Long-Horizon Software Engineering Tasks?SWE-Bench Pro is a challenging benchmark for coding models, featuring complex, enterprise-level problems that require substantial code modifications, with performance evaluations showing significant limitations in current models.Scale AI367 citations21 upvotesRead →Gemma 4 Technical ReportOpen multimodal models from 2.3B to 31B parameters, with a thinking mode and image and audio input.Google DeepMind356 citations81 upvotesRead →Productivity Assessment of Neural Code CompletionFinds that how often developers accept Copilot's suggestions, more than other usage measures, predicts how productive they feel.GitHub349 citationsRead →
Every lab →