Translate
Live intelligence

Latest AI updates

650 updates
AllModelsProductsIndustryResearchTutorialOpinion
Research Editor selected

ToyBlocks: A controlled world for robot learning and evaluation

ToyBlocks is a new simulated environment designed for robot learning and evaluation, offering a controlled world with customizable tasks. It provides a platform for benchmarking and developing robotic manipulation algorithms. The blog post introduces the environment and its features, aiming to facilitate reproducible research in robotics.

Why it matters: This is a concrete, recent AI development that provides a new tool for robotics research, offering practical value to researchers and developers in the field.

Hugging Face Blog4h agoScore 82
robot learningsimulation environmentsbenchmarking
Research

METR audit suggests Anthropic may have crossed its own AI acceleration warning threshold in July

Nonprofit METR audited Anthropic's internal research data over eleven weeks and found that the pace of model development has roughly doubled, with about 80 percent of the speedup driven by AI-assisted research. Anthropic redacted the most informative evidence, so the public must rely heavily on METR's judgment. The findings suggest Anthropic may have crossed its own AI acceleration warning threshold back in July, though some internal voices remain skeptical.

The Decoder4h agoScore 68
AnthropicAI safetyAI development speed
Research

ReferClaw: Disambiguating User Intent for Egocentric Assistance

ReferClaw is an agentic framework for egocentric AI assistants that resolves ambiguous user queries by asking clarifying questions before answering. It uses three cooperating agents to identify candidate interpretations, request clarification, and generate grounded answers, guided by a cost-aware policy that balances information gain against user effort. Evaluations on augmented egocentric QA benchmarks show improved answer accuracy and zero-shot transfer across datasets.

arxiv13h agoScore 75
egocentric assistanceVideoQAuser intent resolution
Research Editor selected

MCPHallu: New Benchmark Exposes Hallucination-Prone Failures in MCP Agents

MCPHallu is a new benchmark designed to diagnose hallucination-prone failures in LLM agents using the Model Context Protocol (MCP). It includes 358 tasks across five domains, each with structural triggers for four failure subtypes: Branch Collapse, Unreachable Goal, Tool Misuse, and Context Forgetting. Evaluation of 14 frontier agents from eight vendors on 5,012 trajectories shows an average score of 0.642, with particular weakness on Unreachable Goal tasks (0.458 average, 23.2% pass rate). The benchmark enables subtype-level analysis, revealing that model rankings shift significantly across subtypes.

Why it matters: Provides a new diagnostic benchmark that goes beyond task completion to identify specific hallucination failure modes in MCP agents, offering actionable insights for improving agent reliability.

arxiv13h agoScore 85
Model Context ProtocolAI agent evaluationhallucination detection
Research

Equivariant Deep Learning in Direct-Product Space for Electronic-Structure Hamiltonian Prediction

Researchers propose DP-Learner, an equivariant learning framework that performs message passing directly on Hamiltonian-like operator blocks in direct-product space rather than direct-sum spaces of irreducible spherical tensors. The method introduces Lie-generator-based directional embedding and invariantly weighted left-right products and contractions intended to preserve equivariance to SO(3), SU(2), and spatial inversions without explicit Clebsch-Gordan decomposition. It is instantiated as a Transformer architecture and accompanied by MP-HAM-110K, a dataset of more than 110,000 material structures covering 68 elements with Hamiltonian and wavefunction data. Experiments on MP-HAM-110K and Materials-HAM-SOC report improved Hamiltonian prediction accuracy and inference efficiency over the state-of-the-art method.

arxiv13h agoScore 77
AI for scienceequivariant neural networkselectronic structure
Research

InfMasking: A Unified Variational Framework for Synergistic Multimodal Learning

InfMasking is a multimodal learning framework that randomly masks each modality's features before fusion, producing diverse partial combinations of cross-modal evidence. The fused representations from masked views must remain consistent with a reference target: an unmasked fused view in unsupervised training or the task-label class proxy when labels are available. The method models masked fused representations as Gaussian and optimizes closed-form variational bounds. On evaluated benchmarks, it improves predictive performance and yields higher mean PID synergy estimates and InterSHAP interaction attribution, with code released.

arxiv13h agoScore 68
multimodal learningvariational inferencefeature masking
Research

Adult-to-Pediatric Fine-Tuning Improves ECG Foundation Model Performance

Researchers adapted an adult ECG foundation model to pediatric ECGs using 339,563 12-lead recordings from 238,189 children, comparing direct transfer, pediatric fine-tuning, and training from scratch. Pediatric fine-tuning achieved a mean AUC of 0.9695, above direct transfer at 0.8256 and de novo training at 0.9661, and reached 0.9079 using only 1% of training data. On 10,999 external ECGs, fine-tuning achieved a mean AUC of 0.9385 across 30 labels and also generalized to age regression and congenital heart disease classification. In a study with 17 clinicians and 100 cases, diagnostic accuracy rose from 56.8% to 65.4% and did not decline when the AI erred.

arxiv13h agoScore 78
ECG foundation modelspediatric cardiologymodel fine-tuning
Research

Token Identity Brier Flow Improves Continued Training of Continuous Language Flows

Researchers propose Token Identity Brier Flow (TIBF), which applies squared supervision directly to token identities when continuing pretrained language flows, combining network scores with token-matching scores under a Gaussian observation model and using the Brier loss for token probabilities. They show that squared error in predicted token probabilities bounds the squared errors of continuous denoising predictions and flow velocities, and provide a risk decomposition separating observation uncertainty from posterior fitting error. Under matched continuation budgets with frozen embeddings, TIBF achieves Gen.PPL values of 33.96 on OWT and 81.17 on LM1B, outperforming continuous-MSE and cross-entropy continuation controls and attaining the lowest Gen.PPL at every tested sampling budget on both datasets.

arxiv13h agoScore 70
continuous language modelingdiffusion modelslanguage modeling
Research

World-Model-Guided Navigation Distills Foresight into Fast Policy

Researchers propose a training-time approach for human-like navigation in multi-agent scenes, using world models as preference oracles to distill foresight into a fast policy. Two task-aligned world models score candidate actions by predicted consequences, and the policy is refined via group-relative preference distillation. At deployment, the world models are removed, enabling single forward-pass action selection. Evaluations across five scenarios show safer, more socially coherent, and more human-like navigation than baselines.

arxiv13h agoScore 74
world modelsrobot navigationreinforcement learning
Research

O²PD: Overlap-Gated On-Policy Distillation Improves Token-Level Alignment

This paper introduces O²PD, a new method for on-policy distillation in language models. It addresses two key questions: which tokens to align and how to align them. The method applies distillation only to the first 20% of student-generated tokens, where teacher supervision is more reliable, and uses an overlap-gated α-divergence to adaptively switch between mode-covering and mode-seeking behavior. Theoretical analysis identifies problems in existing top-k OPD and proposes support strategies to improve alignment.

arxiv13h agoScore 70
knowledge distillationlanguage model trainingalignment
Research

LLM Token Probabilities as Confidence Estimates in Human Behavior Simulation: A Luxembourgish Next Like Prediction Case Study

This paper investigates whether LLM token probabilities can serve as confidence estimates in human behavior simulation, using a Next Like Prediction task on Luxembourgish news platform data. The authors find that open-weight models (Ministral 3B/8B) are overconfident with in-context learning alone, and a simple statistical baseline outperforms them on both accuracy and calibration. Scaling model size improves accuracy but degrades calibration, while task-specific fine-tuning significantly improves both, reducing Expected Calibration Error from 0.2 to 0.006 and boosting accuracy from 0.536 to 0.758.

arxiv13h agoScore 70
LLM calibrationhuman behavior simulationtoken probabilities
Research

RELOAD: Looping Layers in Typed Decision Models Improves Reasoning and Calibration

RELOAD is a method that retrofits loops into pretrained decision models, allowing them to spend variable test-time compute in latent space without adding parameters. When trained with the model, these loops improve accuracy and calibration, especially on problems requiring longer reasoning chains than seen in training. Across six models from 0.4B to 4B parameters, encoders gain up to 23 points, and a lightweight controller preserves most gains at about three-quarters of the compute.

arxiv13h agoScore 74
adaptive computationmodel efficiencyreasoning
Research

Scaling Laws for SAE Training Data

This paper investigates how the number of unique activations affects sparse autoencoder (SAE) training when the total training budget is fixed. It finds that quality improves rapidly with buffer size but then plateaus, where replaying the buffer is nearly as effective as using fresh activations. The authors propose a scaling law for limited data and validate it across various models, layers, and SAE variants. They provide a practical rule: choose the smallest buffer that reaches the quality plateau and replay it, reducing activation storage by 8-64x with negligible quality loss.

arxiv13h agoScore 77
sparse autoencodersscaling lawsmodel interpretability
Research

RIVER: Evidence-guided multi-agent retrieval for enzyme–reaction discovery

RIVER is a multi-agent framework that coordinates LLMs, biochemical tools, and knowledge resources to perform bidirectional enzyme–reaction retrieval. It assembles multi-level evidence and uses validation modules to rerank candidates, achieving top success rates on public benchmarks and outperforming general-purpose and biology-specialized multi-agent systems in E. coli retrieval tasks.

arxiv13h agoScore 72
enzyme function annotationmulti-agent systemsretrieval
Research

New Study Reveals How Irrelevant Context Degrades LLM Accuracy and How to Fix It

A new arXiv paper identifies a mechanism called the Intrusion Pathway that explains why language models lose accuracy when prompts contain irrelevant, off-topic context. The authors localize the problem to a sparse set of 'intrusion heads' and show that knocking out about 1% of attention heads recovers 57% of the performance drop at bounded cost. This offers a practical intervention for improving robustness in RAG, tool-using, and long-context systems.

arxiv13h agoScore 74
language model robustnessattention mechanismsretrieval-augmented generation
Research

Axis-LoRA: Probing and Adapting Embedding Models for Personal-Memory Retrieval

Researchers find that frozen dense embedding models poorly capture time, place, person, and event axes in personal-memory retrieval, with axis changes barely shifting embeddings. They propose Axis-LoRA, a per-axis adapter with four LoRA branches and a per-query gate, which improves retrieval accuracy on benchmarks like LongMemEval, ATM-Bench, and LoCoMo while training only a small fraction of parameters.

arxiv13h agoScore 70
retrieval-augmented generationparameter-efficient fine-tuningpersonal AI
Research

STAG-Net: Lightweight Head Pose Estimation with Structure-Aware Attention

STAG-Net is a new lightweight head pose estimation method that uses structure-aware attention over landmark groups, achieving state-of-the-art accuracy among landmark-based approaches on standard benchmarks. It requires only 0.012 to 0.043 million parameters and 24.1 to 83.6 thousand FLOPs, about three orders of magnitude fewer than FSA-Net. The FaceMesh pipeline runs at 322.9 FPS on a Tesla T4 GPU, nearly twice as fast as the fastest image-based baseline, and training completes within 28 minutes.

arxiv13h agoScore 70
head pose estimationefficient neural networksfacial landmark processing
Research

Masked diffusion as random sequential adsorption: an exactly solvable model for parallel decoding

This paper presents a theoretical framework linking masked diffusion models for text generation to statistical physics of driven many-particle systems. It shows that the error in parallel decoding corresponds to entropy production, and for Markov data, the generation process maps to random sequential adsorption with labelled gaps, solved exactly. The authors derive a thermodynamic metric from mutual information, showing optimal schedules are geodesics, and that longer text correlations concentrate error at the start of generation.

arxiv13h agoScore 75
masked diffusion modelsparallel text generationstatistical physics of machine learning