Machine learning news — deep learning, reinforcement learning, neural architecture search, diffusion models, and new ML frameworks and libraries.
A new arXiv paper (2610.20782v1) argues that open-world AI requires three properties — inference-time learning, a rich representation, and predictive disentanglement — and reports that language models up to 8B parameters…
A new arXiv paper (2610.20748v1) introduces adVersarial Parameter Decomposition (VPD), a method that decomposes a language model's parameters into simple subcomponents each implementing only a small part of the model's l…
A multimodal contrastive learning framework aligned EEG trials with a frozen pretrained text encoder and decoded imagined speech across 39 intent categories with performance comparable to a task-optimized closed-set clas…
IneqAgent, an agentic framework combining LLM reasoning with symbolic computation for inequality proving in Lean4, solved 175 of 186 problems (94.09%) across four benchmarks when run with the DeepSeek-V4-Flash backbone, …
L2G-SR, a framework for LLM-based symbolic regression, achieves state-of-the-art numerical fit, equation-level accuracy and tree edit distance (TED) on LLM-SRBench across the evaluated LLM backbones and domains, accordin…
Researchers released ARIA (Agent Runtime Inspection and Amendment), a framework for learning when and how to repair ongoing LLM-based multi-agent executions, built on a dataset of 3,953 execution-verified repair pairs dr…
Researchers propose EvolvingBrain, an adaptive skill system that lets multi-modal agents self-evolve in open worlds by extracting Procedural and Problematic Patterns from model-specific trajectories and organizing them i…
A new arXiv paper (2610.20805v1) finds that answering from a separate call rather than continuing a model's own reasoning trace raises evidence-following by 29.9 to 60.6 points across gpt-oss, Gemma4 and Qwen3 on HotpotQ…
A compute-aware three-action controller for evaluator-guided test-time reasoning improved observed utility over calibration-selected fixed policies while cutting normalized online compute by about 30% on held-out width e…
A new arXiv paper (2610.20867v1) presents an ASP-based declarative framework for anticipatory commonsense reasoning about interacting motion trajectories, combining an Answer Set Programming model of interactive motion d…
A new arXiv paper (2610.20635v1) compares three diffusion-LLM parameterizations for permutation generation, evaluating them on recommendation reranking across three Amazon domains with candidate sets of size 10 and 20. T…
A team of researchers introduced Temporal Specification Guidance (TSG), a training-free inference-time framework that converts formal Temporal Logic specifications into executable objectives for text-to-video generation,…
Researchers proposed PRISM, a post-training method that improves LLM multi-agent topology generators by learning from privileged sibling graphs, according to arXiv paper 2610.20628v1. PRISM uses a teacher with a zero-ini…
Researchers proposed Competitive Token Selection (CoTS), a training-free framework that dynamically balances discrete token commitment and soft latent exploration at each reasoning step using the probability competition …
A new arXiv paper (arXiv:2610.20615v1) presents DeepUnfolding Agent, a schema-driven FastAPI platform that exposes deep unfolded adaptive control as an external software service for machine learning, federated learning, …
A new arXiv paper (2610.20630v1) presents a corpus of approximately 160,000 Brazilian Portuguese financial news articles collected from web sources to address the scarcity of domain-specific datasets for Financial NLP in…
AbridgeKV, a method detailed in arXiv paper 2610.20641v1, compresses completed blocks of generated text into the KV entries of a few learned pause tokens, allowing a model to keep only a quarter of the generated cache wh…
CoTranslator, a reverse-to-forward chain-of-thought translator from arXiv paper 2610.20828v1, improves average accuracy over semantic suppression by up to 5.82 percentage points across six student models on six mathemati…
Researchers introduced TextDART, a defect-aware region-targeted reward model, and SCANNER, which converts its regional judgments into spatially differentiated policy feedback for visual text generation, per the arXiv:261…
A study of Tagalog tokenization found that a morphology-aware tokenizer, MorphAustro-Tok, beat a corpus-matched BPE baseline by 0.84, 1.49 and 3.40 mean sentence-level chrF++ points on machine translation across 33 runs,…