Translate
Live intelligence

Latest AI updates

1,708 updates
AllModelsProductsIndustryResearchTutorialOpinion
Research Editor selected

Spatial Memory as a Universal Computer: How Finite Brains Generalize Without Bound

This paper proposes that generalization beyond training length arises when a learner reads only coincidences (invariant relations) rather than coordinates (position, count, time). It argues that such a learner's memory must be cyclic, linking to grid cells, place cells, and the hippocampal index, and that this architecture achieves Turing-complete computation. The authors report that their machine trained on strings up to 40 symbols achieves at least 99% accuracy on lengths 41-500 for nine of fifteen benchmark tasks, including five that LSTM/Transformer cannot solve, and can reverse strings of 15,015 symbols when given new cycles.

Why it matters: Offers a novel theoretical framework connecting neuroscience to AI generalization, with concrete benchmark results showing advantages over standard sequence models.

arxiv13h agoScore 82
neural computationout-of-distribution generalizationcognitive architectures
Research Editor selected

ZeroRAG: Decentralized Semantic Routing for Retrieval-Augmented Generation

ZeroRAG is a decentralized routing layer for RAG that enables document retrieval across a peer-to-peer overlay without centralized orchestration or global data structures. Each peer maintains only lightweight local state, and queries are steered via a best-first walk that progressively rewires the network toward semantically relevant neighbors. Evaluations on real-world workloads show it outperforms decentralized baselines in recall and message efficiency, approaches centralized performance, self-heals after failures, and scales to 100,000 peers.

Why it matters: This paper presents a novel decentralized approach to RAG that addresses privacy and custody constraints, with strong scalability and resilience results, offering a practical alternative to centralized retrieval.

arxiv13h agoScore 82
retrieval-augmented generationdecentralized systemssemantic search
Research

AttnSR: Attention Step Reward for Long-Horizon Agentic RL

AttnSR is a step-level reward augmentation method for long-horizon LLM agents that derives reward allocations directly from policy attention during training replay, eliminating the need for a separate reward model or additional rollouts. By tracing backward through the policy model's attention graph, it estimates the relevance of earlier steps to the final response and uses a running baseline to determine whether rewards reinforce or penalize selected steps. In experiments with Qwen2.5-7B-Instruct, AttnSR improved ALFWorld success from 91.2% to 97.1% and consistently outperformed PPO on ScienceWorld while reducing environment interactions by 29.9%.

arxiv13h agoScore 75
reinforcement learningLLM agentsreward shaping
Research Editor selected

CLEAR-WSI: A Scalable Diagnostic-Driven Retrieval Framework for Whole-Slide Images

CLEAR-WSI is a new content-based image retrieval framework for pathology archives that uses fixed-length slide embeddings to enable scalable, diagnostic-driven search. It improves Precision@5 over baseline methods on public and clinical datasets, and supports dynamic expansion of the retrieval memory bank. The framework addresses limitations of handcrafted indexing and predefined sampling strategies.

Why it matters: This paper presents a concrete, validated AI method with measurable improvements on multiple benchmarks, offering practical value for digital pathology and medical imaging retrieval.

arxiv13h agoScore 80
medical imagingimage retrievaldigital pathology
Research

CardioLM: A multilingual suite of small language models for the cardiology domain

Researchers introduce CardioCCC, a corpus of 508 cardiology case reports originally in Spanish and extended to Czech, Dutch, English, Italian, Romanian and Swedish via machine translation, annotation projection and clinician review, annotated for diseases, symptoms, procedures and medications. The accompanying CardioLM suite comprises CardioBERTa encoders built by continued pretraining on cardiology and biomedical corpora, plus CardioNER named-entity-recognition models fine-tuned on CardioCCC. Cardiology-specific pretraining cut pseudo-perplexity by up to a factor of eight but shifted strict span F1 by at most 0.04, while jointly trained multilingual XLM-RoBERTa-large models gave the best strict F1 in nearly every language and entity type, beating the strongest monolingual system by up to 0.13 F1. Clinicians rated CardioNER well above a rule- and dictionary-based baseline, with mean ratings rising from 26% to 73%, though many residual errors were boundary-related, especially for medications.

arxiv13h agoScore 68
CardiologyClinical NLPMultilingual language models
Research

tok: A Textual Concrete Syntax for BPMN/DMN with LLM and MCP Support

The paper presents tok, a plain-text notation for BPMN 2.0.2 process models and DMN 1.5 decisions, with an engine that checks documents against both standards, compiles them to XML with automatic diagrams, and serves language-model applications through the Model Context Protocol. It reports that tok models use fewer tokens than BPMN XML, that its checker detects 98.2% of injected faults versus 85.8% for XML Schema with bpmnlint, and that new language models write valid tok from a one-page reference. Agent evaluations with a lean skill and checker tool reached 97.6% to 100% valid tok across 87 process descriptions, and 44 of 47 simulated event logs fit their source models.

arxiv13h agoScore 73
BPMNDMNModel Context Protocol
Research

ChemE-as-Code: An Open-Source Platform for Agentic AI in Chemical Engineering

Researchers introduce ChemE-as-Code, an open-source environment where LLM-based agents can perform chemical engineering tasks using calculation tools spanning catalysis, process simulation, plant layout, HAZOP and life cycle assessment. Engineering files are machine-readable, versioned in Git, and calculations run outside the LLM, with each value recording its origin and automated checks flagging inconsistencies and out-of-date results by file and line. In an illustrative ammonia case, agents produced a preliminary basic-engineering package for USD 224 of API cost; planted faults and engineering changes showed the checks catch only what their rules cover, so they support rather than replace human review. The design has not been reviewed by an independent engineer.

arxiv13h agoScore 69
agentic AIchemical engineeringLLM agents
Research

Circe: Training-Free LLM Self-Evaluation via Component Interaction Consistency

Circe is a training-free method for judging whether an LLM response is correct without reference answers or repeated sampling. Its internal signals measure directional alignment between earlier attention updates and later attention-induced FFN output changes (component interaction consistency), plus representation geometry along layerwise hidden-state trajectories. A bounded likelihood confidence score from mean negative log-likelihood of generated tokens serves as the external signal, and the three are fused into one correctness score. The authors report that across five LLMs and six benchmarks in four domains it separates correct from incorrect responses better than existing baselines and generalizes to multilingual, multimodal and open-world settings.

arxiv13h agoScore 68
LLM evaluationmodel internalsarxiv research
Research

New Framework Predicts Loss of Control in Frontier AI Systems

Researchers introduce a behavioural framework that decomposes loss-of-control risk into misaligned motive, harm-enabling capability, and monitoring evasion, operationalized via 13 aspects and 29 evaluation tasks. Across 13 frontier models, the resulting LoC score correlates strongly (r=0.84) with failure rates in contained long-horizon scenarios. The framework also guides targeted mitigations that reduce failure rates while preserving general performance.

arxiv13h agoScore 78
AI safetyfrontier AIrisk evaluation
Research Editor selected

TCSExam: A New Benchmark for LLMs in Theoretical Computer Science

TCSExam is a unified benchmark covering theorem proving, solver engineering, constructions, and open conjectures in theoretical computer science. It includes 100 expert-reviewed tasks (Full) and a 20-task Lite version, with evaluations of frontier LLMs like GPT-6 Astra and Opus 5.5. Results show strong alignment between Full and Lite rankings, supporting Lite as a compact discriminative benchmark. The framework provides reproducible evaluation tools and Docker build recipes.

Why it matters: Introduces a rigorous, extensible benchmark for measuring LLM progress on core TCS research, with concrete results from multiple frontier models.

arxiv13h agoScore 85
TCSExamLLM evaluationtheoretical computer science
Research Editor selected

NutriVisRAG: Multimodal Exemplar Grounding for Nutrient Estimation with Vision-Language Models

NutriVisRAG is a retrieval-augmented generation framework that uses a vision encoder to retrieve visually similar annotated meal images as multimodal in-context exemplars for VLM-based nutrient estimation at inference time, without task-specific fine-tuning. Evaluated across 96 configurations (4 datasets, 4 VLMs, 3 retrieval encoders, 2 prompting strategies), it reduced primary estimation error in 56 configurations, with improvements more frequent under whole-meal prompting. For GPT-5.4-mini with whole-meal prompting, the best retrieval encoder reduced error on all four datasets by 1.7–35.7% relative to vanilla prompting.

Why it matters: Presents a novel, empirically validated RAG approach that improves nutrient estimation from meal images without fine-tuning, offering practical value for dietary assessment research.

arxiv13h agoScore 84
NutriVisRAGnutrient estimationvision-language models
Research

DictLLM: Dictionary-Based Post-Training Compression for LLMs

DictLLM is a new post-training compression method that decomposes weight matrices into a shared dictionary and layer-specific sparse coefficients, using SVD and Hessian-based sparsification and quantization. It achieves a 5.1x compression of LLaMA-7B with only a 2.9-point perplexity increase on WikiText-2, outperforming existing low-rank methods. The approach requires no further training and is designed for efficient deployment on mobile devices.

arxiv13h agoScore 72
model compressionlarge language modelsefficient inference
Research Editor selected

MUSS: A Multi-scale and Sequence-based Model for Single-cell Gene Regulation

MUSS is a sequence-based multi-scale model that jointly represents DNA sequence, chromatin accessibility, and gene expression at single-cell resolution. It uses a frozen pre-trained sequence encoder and integrates peak and gene representations via self-attention and distance-biased interactions. The model achieves strong performance on gene expression prediction, enhancer-gene linking, and zero-shot cell clustering.

Why it matters: Introduces a novel sequence-grounded multi-scale representation approach that outperforms baselines in single-cell regulatory modeling, with practical code and data release.

arxiv13h agoScore 83
gene regulationsingle-cell omicsdeep learning
Research

Who Can Evaluate AI Independently? A Framework for Analysing Evaluation Infrastructure

This paper introduces PAAI (Power, Access, Adequacy, Independence), a framework for analyzing who controls AI evaluation infrastructure and how that shapes independent evaluation. Applying it to African AI evaluation projects and joint testing exercises of AI Safety Institutes, it identifies patterns of extractive collaboration, dependent innovation, and constrained independence. The findings show that participation in evaluations does not automatically confer authority over design, execution, or resources.

arxiv13h agoScore 70
AI evaluationAI governancepower dynamics
Research

PCBRouteAgent: LLM Agent for Automated PCB Routing

PCBRouteAgent is a multimodal PCB routing agent that combines a Global Routing Planner with a Detailed Routing Executor to automate printed circuit board routing. It uses board images and structured electrical/geometric data to guide KiCad's native routing engine, with bounded local recovery and structured feedback for revisions. Evaluated on 247 boards across four PCBWorld-Bench splits, it achieves clean-pass rates of 100%, 94.9%, 100%, and 40%, outperforming baselines on real-world subsets by 10-36 percentage points.

arxiv13h agoScore 72
AI agentsprinted circuit board designmultimodal reasoning
Research

TOPSD: A New Method to Prevent Teacher Drift in Multilingual Reasoning Distillation

Researchers propose Teachable On-Policy Self-Distillation (TOPSD) to address Teacher Drift in multilingual reasoning distillation, where the teacher model follows student errors instead of correcting them. TOPSD identifies where drift begins using teacher-student likelihood and only applies supervision to prefixes that pass verification. Experiments across ten languages and three model scales show consistent improvements over conventional OPSD, especially in low- and medium-resource languages.

arxiv13h agoScore 70
self-distillationmultilingual reasoningteacher-student learning
Research

Study of 489K LLM Conversations Reveals Gap Between Grounding Claims and Inspectable Evidence

An analysis of 489,283 assistant-user response pairs from ShareChat and WildChat finds that broad attribution language appears in 31.9% of responses, while explicit URLs appear in only 1.33%, highlighting a gap between claims of grounding and preserved inspectable support. The study examines how response cues like attribution and uncertainty relate to users' accepting, checking, or challenging behaviors, finding that uncertainty is the only cue with a small statistically distinguishable association favoring checking. Scrutiny is concentrated in certain conversations, and its turn-to-turn persistence diminishes when conversation-level differences are accounted for.

arxiv13h agoScore 72
Large language modelsUser trustEvidence grounding
Research

BEEMO: Decoupled Behavioral Pretraining Scales Neural Motor Decoding

Researchers introduce BEEMO, a multi-session framework that decouples behavioral and neural representation learning for neural motor decoding. It trains a behavior decoder on abundant unpaired behavioral data and incorporates neural activity via cross-attention when paired recordings are available. Experiments on nonhuman primate datasets show that adding unpaired behavioral data improves decoding performance, especially for transfer to new tasks, and behavior-only pretraining matches models pretrained on larger neural datasets.

arxiv13h agoScore 75
neural decodingbehavioral datapretraining