AI Research news and analysis on Web Pulse: 22685 curated articles tracking the latest AI Research developments, tools, and research, updated continuously from vetted sources.
Researchers proposed X-CoSD, a cross-vocabulary collaborative speculative decoding framework for distributed LLM inference that works with heterogeneous small language model and server LLM vocabularies, according to an a…
A new arXiv paper (2609.05461v1) introduces ARC-Bench, a no-leak, fixed-candidate protocol that finds frozen JEPA-style world models rank candidate actions incorrectly, with the top-scored candidate almost always subopti…
A new arXiv paper introduces SWORD (Systematic Wikidata-based Object-Relation Distortion), a benchmark that tests whether large language models consistently reject factually incorrect statements across eight widely spoke…
Noora Health's nurses, who handle more than 50,000 medical queries per month on a WhatsApp-based caregiver support service, raised emergency triage recall from 0.565 to 0.810 and F1 from 0.606 to 0.702 by decomposing the…
A study posted to arXiv (2609.09363v1) found only a weak context-memory conflict in large language models, with counterfactual RDF triple inputs scoring just 0.05 lower on a 1-5 faithfulness scale than factual inputs whe…
Snowflake AI Research introduced HybridDeepResearch, a benchmark of 380 tool-dependent tasks that require agents to combine web search and SQL against LiveSQLBench-Base-Lite databases and public web corpora. Evaluations …
Researchers released BuzzASR, a collection of 102 language-specialized fine-tuned Whisper models for automatic speech recognition, according to an arXiv paper (2609.09554v1). BuzzASR models outperform Whisper-large-v3 on…
Researchers introduced Edu-QuRating, a pipeline that scores and curates educational text on multiple education-specific criteria by distilling LLM-judge pairwise preferences into reusable Edu-QuRaters, with the best scor…
Researchers proposed EvoTree, a staged framework that generates evolution trees from citation graphs by decoupling conceptual backbone learning from temporal refinement, according to the arXiv paper 2609.09561v1. EvoTree…
Researchers introduced jaROTE, a rule-based pipeline that reproduces omitted temporal expressions in Japanese news articles as concrete dates or intervals using the publication date as external context before indexing fo…
Researchers introduced MonoTM, an interpretable topic modeling framework that decouples document-topic mixture estimation from semantic interpretation, according to a paper posted to arXiv as 2609.09575v1. MonoTM estimat…
A new arXiv paper (arXiv:2609.05459v1) proposes a deterministic framework that compiles games specified in the Video Game Description Language (VGDL) into Dynamic Structural Causal Models, directly translating sprite dyn…
Researchers introduced EnvCraft, an automated framework that synthesizes executable environments and training data for Agentic Reinforcement Learning, detailed in arXiv paper 2609.05576v1. EnvCraft's environment synthesi…
Researchers studying Looped GPT-BERT in the BabyLM 2026 Strict-small setting trained a 12.18M-parameter model on a preprocessed 7.48M-word English corpus, using four physical layers for twelve recurrent traversals, and t…
Google Gemini 3.0 Pro recovered only 2.8% of a 180-contaminant answer key when auditing large batches of documents, down from 50% on single documents and 60% on small batches, according to an arXiv paper (2609.09696v1) t…
A new arXiv paper (2609.05437v1) introduces NormReact, a dataset of 450 hand-annotated norm violation scenarios, to evaluate second-order social reasoning — metanorms — in large language models. Across six models, the re…
A new arXiv paper (2609.05448v1) formulates structured post-training pruning of vision and language transformers as a damage-aware multi-armed bandit problem under a fixed candidate-evaluation budget, masking attention h…
A reasoning-aware compression framework that selectively restores the most quantization-sensitive circuits to FP16 achieves Pareto-optimal energy-accuracy points unreachable by uniform quantization, according to an arXiv…
A World Values Survey-grounded simulation of LLM agent societies found that more than 50% of personas fail to express their assigned WVS value profiles from the outset, while only 2-7% drift after repeated conversations,…
Researchers proposed TEAM-Design, a rule that assigns every task two replay probabilities — one per baseline — to decide under a fixed replay budget whether a human-AI workflow should be kept over the human alone or the …