llm

RSS

61
Designer-RSI: Evolving Procedural Memory from User Traffic for Agentic Graphic Design arxiv.org
The paper introduces Designer‑RSI, a continual‑adaptation system that couples a frozen design‑software‑controlling LLM with an external procedural memory of natural‑language design skills. Over five adaptation rounds on real user briefs, the memory grows from 76 to 139 procedures and lifts execution success from 72.7% to 99.3%, showing that skill accumulation and selective replay can dramatically…
Hugging Face Daily Papersarxiv.orgHongyang Du, Lan Yan, Christian Flores4d1 minpaper
62
Towards Self-Driving Codebases detail.dev
The post argues that AI agents could eventually handle low‑level engineering tasks—bug fixing, debugging, UI consistency, growth experiments—if the dev toolchain is made “agent‑legible”. It outlines missing primitives (global memory, code‑base rot prevention, better dev environments) and proposes a bootstrapping process to measure and improve a repo’s “agent readiness”. The piece is largely specu…
Hacker News front pagedetail.dev5d9 minHN12099
63
How do we prevent mathemathics from devolving into the Medieval Era of secrecy?mathoverflow.net
A MathOverflow discussion argues that AI‑driven proof generation threatens the open, collaborative nature of mathematics, proposing human‑only attribution and mandatory disclosure of AI methodology as possible safeguards.
Hacker News front pagemathoverflow.netSteven Landsburg, Qfwfq, David White, R.P., D. Corro, Stanley Yao Xiao, Gerry Myerson, Alexander Woo, Graham7d12 minHN157146
64
RecreationWorld: Scalable and Verifiable Environments for Hybrid Computer-Use Agents arxiv.org
The paper presents RecreationWorld, a five‑platform framework that lets hybrid computer‑use agents learn by recreating the behavior of a running reference, and introduces RecreationBench, a 250‑task benchmark with programmatic and visual assertions. Experiments show GPT‑6 Astra reaches 58.1% overall but struggles with deeper programmatic tests, highlighting gaps in current agents.
Hugging Face Daily Papersarxiv.orgShuai Bai, Jiayong Deng, Yikun Fu4d2 minpaper
65
Benchmark Radar: A Living Database and Search Engine for AI Benchmarks and Evaluation arxiv.org
Benchmark Radar is a continuously updated database and search engine that aggregates AI benchmark papers, datasets, and scores from 37 sources into a searchable catalog. It offers a web dashboard, CLI, and analysis of benchmark saturation to help LLM developers find and compare evaluations.
Hugging Face Daily Papersarxiv.orgKoutian Wu, Junjie Zhou, Ergan Shang12d1 minpaper
66
Show HN: Share your AI Setup, Learn from others mysetup.ai
A Show HN post announcing mysetup.ai, a community‑focused site where users can list the LLM tools, agents, and workflows they use. The author shares a few example profiles and invites others to add theirs, but provides no technical details, design rationale, or measurable insights.
Hacker News front pagemysetup.ai5d1 minHN244138
67
ActionPiece: Rethinking Action Tokenization for Autoregressive Vision-Language-Action Models arxiv.org
Action tokenizers for autoregressive VLA models often fail to preserve subtle action adjustments, despite good pointwise reconstruction. This paper introduces Physical Rank Consistency (PRC) to measure relational fidelity and ActionPiece, a new tokenizer that uses joint supervision to preserve these physical relationships, significantly improving policy success on robotics benchmarks.
Hugging Face Daily Papersarxiv.orgShijie Lian, Bin Yu, Zhaolong Shen6d1 minpaper
68
Grounded Skill Synthesis from Code at Scale for Agentic Intelligence arxiv.org
Code2Skill is an automated pipeline that mines popular GitHub repositories to extract verifiable, implementation‑anchored procedural “skills”. It builds a bank of ~1 M skill records (atomic ops, workflows, patterns) with provenance metadata, verifies each via blind reconstruction, and shows that augmenting LLM‑based agents with these skills yields an average 11.7% performance lift across 72 proto…
Hugging Face Daily Papersarxiv.orgYongqi Tong, Pan Wang, Hang Wang18d1 minpaper
69
A Zeroth-Order Paradigm for LLM Preference Alignment arxiv.org
The paper proposes Comparison-based Preference Optimization (ComPO), a zeroth‑order method that uses comparison oracles to align LLMs without a differentiable loss. Experiments on several LLM families show it improves win rates and mitigates likelihood displacement compared to direct alignment approaches.
Hugging Face Daily Papersarxiv.orgPeter Chen, Xi Chen, Wotao Yin6d1 minpaper
70
Fathom: Per-Query Read Depth for Sparse Decoding over Offloaded KV Caches arxiv.org
Fathom introduces a per-query read depth mechanism for sparse decoding over offloaded KV caches, allowing each query to adaptively decide how many bits of each key channel to read. This method significantly speeds up decoding for large language models with long contexts by reducing host memory traffic, achieving 1.67x faster GPU decoding on Qwen3-8B at one million tokens.
Hugging Face Daily Papersarxiv.orgVivek Kalyanarangan5d1 minpaper
71
BI-Agent and BI-Bench: Towards Automating End-to-End Business Intelligence arxiv.org
The paper introduces BI‑Bench, a new benchmark of real‑world BI questions derived from public dashboards, and BI‑Agent, a tool‑augmented LLM system that breaks BI workflows into search, join, and transform subtasks. Baseline LLMs hit <50 % accuracy on BI‑Bench. By orchestrating specialized data‑management tools and post‑training the model with supervised fine‑tuning and reinforcement learning on…
Hugging Face Daily Papersarxiv.orgChuxuan Hu, Yeye He, Penny Zhou, Wee Hyong Tok, Daniel Kang, Surajit Chaudhuri1d2 minpaper
72
MintAct: A Unified Visual Agent for Digital Environments arxiv.org
MintAct is a family of vision-language models (2B-8B) that unifies UI grounding, multi-step navigation, and visual tool use across mobile, desktop, and web environments. It achieves state-of-the-art performance by leveraging a scalable environment and an asynchronous reinforcement learning infrastructure.
Hugging Face Daily Papersarxiv.orgMingfei Gao, Rui Tian, Haiming Gang4d1 minpaper
73
The Last AI Built by Humans: Toward Genuine Recursive Self-Improvement arxiv.org
The paper defines recursive self‑improvement (RSI) for AI, introduces the Headroom‑Closed Index to expose current LLM limits, and proposes a staged roadmap toward full RSI. It surveys application scenarios and outlines practical challenges.
Hugging Face Daily Papersarxiv.orgYi Duan, Ying Liu, Zirui Tang12d1 minpaperHN3
74
OmniVChat: Synthesizing, Benchmarking, and Training for Native Audio-Visual Dialogue arxiv.org
OmniVChat defines native audio‑visual dialogue where a model consumes raw audio and video streams and replies in text. The authors build OmniVChat‑Studio, a multi‑agent simulator that generates single‑ and multi‑turn audio‑visual conversations, and use it to create OmniVChat‑Bench, a benchmark covering five dialogue abilities. They also propose OmniVChat‑RL, a reinforcement‑learning reward that b…
Hugging Face Daily Papersarxiv.orgHaolin He, Yunfei Chu, Qi Chen4d2 minpaper
75
Show HN: Scry, programmable internet search w/ congestion pricing scry.io
Scry is a read‑only, SQL‑like search service that lets LLM agents query billions of public‑web records (e.g. Reddit, Hacker News, academic papers) via a single MCP or HTTP endpoint. It enforces per‑query limits (deadline, memory, row cap), offers schema discovery, explain, rerank, and compile APIs, and reports freshness per source (e.g. Hacker News <15 min). In a DeepSearchQA benchmark Scry‑enabl…
Hacker News front pagescry.io4d28 minreleaseHN6025
76
Confidence Comes from Experience: Experiential Confidence Estimation from Reasoning to Agents arxiv.org
XConf (eXperiential Confidence) is a novel method for estimating language model confidence by leveraging the model's accumulated experience from past graded episodes. It significantly outperforms existing methods like self-consistency in discrimination and calibration, at a fraction of the computational cost, across various tasks.
Hugging Face Daily Papersarxiv.orgCaiqi Zhang, Xiaochen Zhu, Chengzu Li7d2 minpaper
77
Self-Evolving Search Index arxiv.org
The paper introduces SELF-INDEX, a framework that lets a search index automatically diagnose retrieval failures, revise its keys, and validate changes, using a query simulator to anticipate future queries. Experiments show consistent gains across corpora and downstream LLM agents.
Hugging Face Daily Papersarxiv.orgSangam Lee, Wonjae Lee, Sunghwan Kim5d1 minpaper
78
When2Think: Learning Difficulty-Aware Length Control for Efficient Hybrid Reasoning Models arxiv.org
When2Think introduces a post‑training framework that lets a large reasoning model decide per‑instance how much reasoning depth to allocate, using difficulty‑aware reward shaping (IDAC) and verifier rewards. It cuts token usage by ~28% while boosting Pass@3 by 10% on AIME24 and reaches 40% Pass@3 on AIME25, outperforming compression and routing baselines.
Hugging Face Daily Papersarxiv.orgJaejun Shim, HyunJin Kim, Young Jin Kim5d1 minpaper
79
Show HN: How Stale Is Your AI? Release age and training cutoff for 20 models jock.pl
The author curates release dates and training‑cutoff dates for 20 recent LLMs, exposing how many months each model lags behind its launch. He also provides JSON/txt files and instructions for agents to fetch this data to avoid stale assumptions.
Hacker News front pagejock.plPawel Jozefiak6d4 minHN7949
80
The Other Half of the Memory Wall: Serving 35B MoEs from SSD with Trained Routing Prediction arxiv.org
The paper presents Edge0, a streaming MoE inference engine that predicts the next layer's routing one token ahead, allowing expert weights to be fetched from SSD while compute proceeds. This enables a 35 B‑parameter MoE to run on a single 24 GB machine at ~20 tokens/s using only ~3 GiB of active memory and with near‑teacher accuracy.
Hugging Face Daily Papersarxiv.orgYu Lin, Yiming Wang, Runyuan Cai6d1 minpaper
81
Sample Count Is Not Enough: Candidate-Generation Strategy Shapes the Energy and Performance of LLM Test-Time Scaling arxiv.org
Increasing the number of LLM candidates (N) improves reasoning accuracy, but the way those candidates are generated (batch size vs sequential calls) dramatically affects latency, GPU‑hours, and energy. On A100 GPUs, eight serial 1‑candidate calls consume ~5× more energy and take ~6× longer than a single 8‑candidate batched call, even though total candidate count is identical. The authors recommen…
Hugging Face Daily Papersarxiv.orgMobina Kashaniyan, Ali Jannesari4d2 minpaper
82
Markdown in /src htmx.org
The essay argues that Markdown should be treated as source code and checked into a /src/md directory alongside the implementation it describes. Keeping intent in diffable, reviewable Markdown lets teams generate code and tests from it while preserving locality for both humans and AI agents.
Lobstershtmx.org16h6 minlobste.rs10
83
VABench: Measuring Embodied Spatial Intelligence through Visual Demonstrations, Active Perception, and Metric Control arxiv.org
VA‑Bench is a new benchmark that evaluates general‑purpose multimodal LLMs on the full observe‑reason‑act‑revise loop in embodied robotics, using RGB demonstrations, active camera control, and metric Cartesian commands. The best model reaches 53.9% average task success, showing active perception helps but long‑horizon tasks remain unsolved.
Hugging Face Daily Papersarxiv.orgZhongbo Zhang, Jiayi Jin, Yifan Wang5d1 minpaper
84
StepAudio 3 Realtime Technical Report arxiv.org
StepAudio 3 Realtime is an audio‑language foundation model that runs a continuous listen‑converse‑think‑act loop. It introduces Deep Perception for rich acoustic cue extraction, Seamless Duplex for handling pauses/back‑channels, and a Think‑While‑Speaking mechanism that lets the model reason in parallel with speech output. On benchmarks it scores 73.0 macro avg on StepAudioChat, 90.6 on MMSU, 98.…
Hugging Face Daily Papersarxiv.orgBin Lin, Bo Zhao, Boyang Zhang10d2 minpaper
85
Heretic removes restrictions from language models heretic-project.org
Heretic is an open‑source Python package that claims to strip safety or policy restrictions from LLMs so they obey any prompt. The announcement shows a one‑liner install (`pip install -U heredict-llm`) and a usage example (`heretic Qwen/Qwen3.5-4B`). Links to GitHub, Hugging Face, Discord and Matrix are provided, but no technical details, design rationale, benchmarks, or code snippets beyond the…
Hacker News front pageheretic-project.org1d1 minreleaseHN19380
86
RSA-896 saweis.net
Stephen A. Weis factored the RSA-896 challenge number using Claude on September 19, 2026. The post provides the prime factors p and q for the 896-bit number.
Hacker News front pagesaweis.net2d1 minHN22590
87
What Does Privileged Information Add to On-Policy Self-Distillation?arxiv.org
The paper isolates the contribution of privileged teacher information in on‑policy self‑distillation (OPSD) using a 5k‑problem math suite. Reference‑free distillation explains most gains, with only modest extra benefit from full solution traces.
Hugging Face Daily Papersarxiv.orgXiuYu Zhang, Wei Chow, Junfeng Fang5d1 minpaper
88
Verifiable Social Reasoning for LLM Assistants arxiv.org
The paper introduces Fuse, a multi‑agent simulation that gives LLM assistants a verifiable ground‑truth task for social reasoning by hiding a target agent’s motive and letting a user‑mediated conversation infer it. Experiments on 12 LLMs show user mediation makes reasoning harder, models are biased by user framing, need more detail than humans, and longer chats don’t always help.
Hugging Face Daily Papersarxiv.orgAmir Taubenfeld, Zorik Gekhman, Avigail Grinstein-Dabush7d1 minpaper
89
Vibe Coding is the new Internet Dating?joecmarshall.com
A speculative opinion piece comparing the rise of AI‑assisted "vibe coding" to the evolution of internet dating, arguing that as AI becomes the default coding partner the term will become obsolete.
Hacker News front pagejoecmarshall.comJoe Marshall7d2 minHN6299
90
HypoEvolve: Genetic Algorithms Enable Multi-Agent LLMs to Discover Scientific Hypotheses arxiv.org
The paper presents HypoEvolve, a generational genetic algorithm that coordinates specialized LLM agents to iteratively propose, critique, and refine scientific hypotheses. On a drug‑repurposing benchmark across 34 cancer types, it outperforms six baselines, achieving a DepMap selectivity of 0.171 versus 0.115.
Hugging Face Daily Papersarxiv.orgJieyuan Liu, Mengzhou Hu, Jefferson Chen8d1 minpaper