与「authors」相关的搜索结果

RAG system that skips HTML parsing entirely! PixelRAG is an open-source visual RAG framework that renders documents as screenshots instead of parsing them into text. Most RAG pipelines start by converting HTML to text. Tables flatten into unstructured rows. Charts disappear. Layout context is gone before the LLM ever sees it. The paper measured this directly: HTML-to-text conversion accounts for 36.6% of retrieval failures on SimpleQA. PixelRAG skips that step entirely. It renders pages as screenshot tiles using Playwright, embeds those tiles with a fine-tuned Qwen3-VL-Embedding model, builds a FAISS index, and passes retrieved images directly to a VLM reader. No text abstraction in between. Benchmarked across six datasets against the strongest text-based baselines: - SimpleQA: 78.8% vs 71.6% (+7.1 points) - NQ-Tables: 48.8% vs 42.5% (+6.3 points) - EVQA: +15.5 points - LiveVQA: +11.3 points One honest caveat from the authors: this requires Qwen3-VL-4B class models or larger to see the benefit. Smaller models trail text retrieval. The authors also recommend using PixelRAG as an enhancement layer alongside existing text systems rather than a full replacement. Ships with a pre-built Wikipedia index covering 8.28M articles across 28.1M screenshot tiles. A Claude Code plugin lets Claude take screenshots of any URL and reason over the visual content directly. Key capabilities: • Renders web pages, PDFs, and images as screenshot tiles via Playwright • Fine-tuned Qwen3-VL-Embedding model for visual retrieval • FAISS index for fast vector search • Pre-built Wikipedia index: 8.28M articles, 28.1M tiles • 3x token cost reduction via image compression • Claude Code plugin for direct URL screenshot and visual reasoning • LoRA fine-tuning support via pixelrag-train 100% open source. I've shared the link in the replies!

🚨 Thrilled to share that our lab will be presenting the 🏆 Best Paper at the NExT-Game Workshop at #ICML2026# today! 🎤 When Agents Lie: Premeditation, Persistence, and Exploitation in Repeated Games 🏆 Best Paper @ NExT-Game Workshop 📍 Conference Room S307 📅 Fri, Jul 10 🕐 13:00–13:20 KST Authors: @JerickShi @TerryJCZhang @bschoelkopf @conitzer @ZhijingJin 🤖 We introduce a three-stage endogenous promise protocol for repeated multi-agent games that asks not only whether LLM agents honor their public commitments when they can privately deviate, but also how model-on-model composition influences premeditated deception and persistent exploitation. 📊 Across six canonical games spanning binary and numerical action spaces, our evaluation of frontier models (GPT-5.2, Llama-4-Maverick, Claude-Opus-4.6) reveals: 🔹 Over 90% of promise-breaking instances are premeditated in agents' private plans. 🔹 Mixed-model groups with mismatched communication frameworks create systemic, persistent payoff gaps of up to 5.00 points from Round 0. 📄 Paper: #MultiAgentSystems# #LLMs# #GameTheory# #AI# #ICML2026#

📣 We are presenting 6 main conference papers 🚀and 14 workshop papers (including 🏆2 Best Papers🏆) at #ICML2026# in Korea! Also hosting one of the largest workshops, Trustworthy AI for Good, on July 10th 🌍❤️. We push the frontiers on #AISafety#, #MultiAgent#, and #CausalReasoning# at @JinesisLab! 🎉 Huge congratulations to all collaborators and co-authors. Excited to discuss these projects in Seoul! Feel free to reach out and talk to our 20+ members and collaborators in Korea @ZhijingJin, @_AndreiMuresanu, @iarthsingh, @ChanglingXavier, @davidguzman1120, @EmanuelTewolde, @ettogran, @FurkanDanismann, @Jerick1380, @PepijnCobben, @rishit_dagli, @_rfaulk, @SimkoSamuel, @TerryJCZhang, @vantru0ng, @zhxiao03, @x_angelohuang, @yahang_qi, @ozzaney0101, @_yongjinny. Happy for collaboration on any of the above topics 🤝 EuroSafeAI, University of Toronto, ETH Zürich, Max Planck Institute for Intelligent Systems Main conference spotlight 🌟 Safe Models Do Not Guarantee Safe Societies: The Case for Sociopolitical Risk Main conference posters 📌 CauSciBench: Can LLMs Automate Causal Inference in Real-World Scientific Research? 📌 Training with Honeypots: Reshaping How LLMs Fail 📌 CoopEval: Benchmarking Cooperation-Sustaining Mechanisms and LLM Agents in Social Dilemmas 📌 Trustworthy AI Suffers from Invariance Conflicts and Causality is The Solution 📌 LLM for Physics Research Requires Domain-Specialized Training and Tooling Workshop best papers 🏆When Agents Lie: Premeditation, Persistence, and Exploitation in Repeated Games BEST PAPER@NExT-Game Workshop 🏆Transferability for General Reasoning: An Automated Curriculum for Multi-Domain LLM RL BEST PAPER@RLxF Workshop Workshop oral and spotlight 🎤 AF-ARENA: A Multi-Dimensional Evaluation Suite for Alignment Faking — AIWILD 🌟 Multi-Agent AI Systems Need Institutional Design, Not Just Model-Level Alignment — AI4GOOD Workshop papers 📄 The Wedge Questions: Latent Cultural Boundaries in LLMs via Persona Projection Divergence — Pluralistic Alignment 📄 GT-HarmBench: Benchmarking AI Safety Risks Through the Lens of Game Theory — NExT-Game 📄 Stargazer: A Scalable Model-fitting Benchmark Environment for AI Agents under Astrophysical Constraints — AI4Physics 📄 Test of Time: Rethinking Temporal Signal of Benchmark Contamination — FoGen 📄 CoopEval: Benchmarking Cooperation-Sustaining Mechanisms and LLM Agents in Social Dilemmas — AI4GOOD 📄 Weight-Level Defenses Improve LLM Agent Adversarial Robustness — AI4GOOD 📄 Evaluating Cooperation in LLM Social Groups through Elected Leadership — AI4GOOD 📄 Causal AI Scientist: Towards End-to-End Causal Inference with Large Language Models — AI4Research 📄What Game-Theoretic Benchmarks Miss: Strategic Silence in Multi-Agent LLMs — FAGEN 📄Proving Your Way to Cooperation: Formalizing Proof-Based Open Source Game Theory in Lean — AI4Math

Beautiful paper from Google DeepMind. Explains the pathways from AGI to ASI, and why that jump could happen through several routes. The authors frame the AGI-to-ASI transition around 4 technical pathways: - continued scaling of compute, model size, data, and test-time inference; - algorithmic paradigm shifts beyond today’s transformer-based foundation-model stack; - recursive self-improvement, where AI accelerates AI R&D and improves future systems; and - multi-agent collective intelligence, where large populations of specialized agents coordinate into a superhuman group agent. Scaling may work for a while, but it could hit limits in data, compute, energy, or weaker returns from making systems larger. Recursive improvement is the most uncertain path, because AI could speed up AI research, but that loop may also slow if hard research problems need real-world testing, scarce hardware, or new ideas. Multi-agent collectives may be the most underappreciated path, because a society of competent digital workers could outperform a brilliant individual model through specialization, speed, and coordination. The big point is that ASI may not arrive as 1 sudden event, but as a chain of faster changes as AI helps create better AI and stronger scientific tools. ---- – arxiv. org/abs/2606.12683 Title: "From AGI to ASI"

❓️ PROTEUS Trial - thoughts? A Letter to the Editor & an editorial pubished in EUO ➡️ PROTEUS: A Practice-changing Trial, but Are We Ready to Embrace Perioperative Apalutamide for All High-risk Patients? by Giuseppe Di Lorenzo & Francesco Montorsi @profmontorsi ➡️ PROTEUS Trial the Day After: Practice Changing or Premature Escalation? by Isabel Heidegger...Pierre Blanchard et al @heid58187 ➡️ Thank you to the authors for sharing their valuable insights! @uroweb @mroupret @GPloussard @jteoh_hk @Ric_Campi @CaPsurvivorship @LauraMarandino @RenuEapen @Ecastromarcos @OncoAlert @Sciencedirect

We are open-sourcing blcli: an Agentic Infra Stack, battle-tested at 30M+ user scale. It allows coding agents like Codex or Claude Code to help manage your whole cloud infrastructure through code, PRs, dry-runs, and deterministic apply workflows. A solid & serious infra that can support to millions of users. This is a collaboration across multiple teams, the same stack that powers @AlvaApp, @Galxe, @GravityChain, and @ReahPlatform. Check it out here: Docs: blcli: Production stack template: Personal account starter: A common take today is: AI agents are useful for toy apps and prototypes, but not for serious infrastructure. The conclusion is wrong, because the issue is not that agents cannot work on real systems. The issue is that real infrastructure requires a large amount of expert context to get it correct in the first place, and even more context to guide agents through the next 18 months of iteration. Production infrastructure is not just a few Terraform files or Kubernetes YAMLs. It includes: cloud projects IAM boundaries networking VPC / subnet / firewall design Terraform state and backend management Kubernetes clusters cluster add-ons secrets management Git-based deployment workflows observability and telemetry (logs, metrics, traces. All integrated together and ready for your Agents to debug live on your prod env) databases, often self-hosted for cost efficiency and control environment separation: stg / beta / prd operational runbooks rollback paths production failure patterns Most of this knowledge usually lives in senior engineers’ heads, internal docs, shell scripts, Slack threads, old runbooks, and lessons learned from real incidents. If an agent does not have that context, of course it will build toy infrastructure. So the real question is: How do we package production infrastructure expertise into a form that AI agents can read, reason about, modify, and operate safely? That is what blcli does. At its core, blcli is a CLI tool plus a whole package of best practices of Infrastructure as Code. The key design principle is simple: Agents are already very good at reading and modifying code. So we make infrastructure code-first. The generated repo is intentionally self-explanatory. An agent can open the repo and understand what happened, and what's next. Who blcli is for? We built blcli for two types of users. 1. Product teams that need to scale beyond prototypes The first group is teams building real products that need infrastructure capable of growing beyond the prototype stage. These teams want the speed of AI-assisted development, but they cannot afford toy infrastructure. 2. Frontier labs and agent teams building self-improving systems The second group is frontier labs, data companies, and agent teams that need infrastructure not just to run applications, but to train, evaluate, and improve agents. If you are building coding agents, infra agents, or long-horizon autonomous systems, blcli stack is a good agent harness/env. Authors: @SiriJhui @p0pUBhv35I8308 @alvinFu1 @ryan4yin @algoxstonk

Might be a hot take, but I didn’t read the @jpmorgan blog post as backing the Clarity Act, as some have suggested. While the authors praised digital assets for their potential to enable tokenization and modernize financial infrastructure, they stopped short of endorsing any specific legislation, instead advocating more broadly for a digital asset market structure “framework” with stronger safeguards, while echoing concerns JPMorgan CEO Jamie Dimon has previously raised about stablecoin yield and illicit finance. The line that stood out to me was the paragraph immediately after the authors outlined the bank’s digital asset initiatives: “Responsible innovation is already possible within existing guardrails, and it can scale further with the right framework in place.” To me, that suggests JPM believes it can already innovate responsibly under today’s regulatory framework, while viewing market structure legislation as beneficial if it also addresses long-standing concerns around stablecoin yield and illicit finance.

Michael Crichton is such a fascinating figure and it is a goddamn shame he died so young This guy got into Harvard Medical School and said "actually, I want to write stories" so he wrote one of the most harrowing long short stories, the Andromeda Strain. This guy was addicted to research. He read scientific journals for fun and researched everything that captured his interest. He was curious and skeptical. He thought hard about the implications of technology all the time. And, instead of being a dork and writing a blog, he wrote entire novels warning about the dangers of emerging technologies. And they were great! He wrote a `Prey` about the dangers of combining AI with nanobots 24 years ago. It's tremendous. You could publish it today and it would be relevant. I miss him. I miss technically competent authors who can spin a good yarn while informing the reader about how technology is changing the world. Crichton was a generational talent and we are poorer for his absence.

I was invited to review for NeurIPS 2026 and assigned 4 papers. 2 of 4 are complete AI slop and deserve strong rejects. They study homework-level problems, use overly lengthy yet uninformative language, and include obviously AI-generated method figures that are not even vector graphics. I just found an interesting shortcut for identifying these papers: check the number of references. The authors obviously did not systematically survey the area, and many of these papers cite fewer than 20 works in total.

New studies drawing on millions of English-language scientific articles by Chinese authors suggest they were largely ignored by Western researchers. That is not for lack of value