Search: “papers”

If you use LLM-as-judge, this one is worth reading. (bookmark it) It's actually one of the most effective ways to use LLM-as-a-Judge for evals. Holistic judge scores hide both their reasoning and thei

Thanks for running our open-source work on current frontier models “The results are: the most capable models today (GPT-5.5 Pro) did outperform the best models from before (79/100 vs 69/100), but did

The take that frontier models aren't ready for use in medicine is dead wrong. I've been using Opus 4.x in my clinical workflow every day for 6 months. I'm a deep sub-sub-specialist in dermatology - th

New research from Meta. Building synthetic training data has stayed a fixed pipeline that you hand-tune and then freeze. Autodata casts an AI agent as a data scientist that builds training and evaluat

This has a lot in common with how controlled self-evolution works in Row-Bot. And that built on top on LangGraph too. Here's an article I wrote about how it works a few days ago: https://x.com/i/statu

Video models are hard. Real-time video models are even harder. Loved this breakdown by @itunpredictable on how we productionized Runway Characters, our real-time interactive avatar model. We had to fi

📣📣 Meet Qwen-AgentWorld — a native language world model that simulates 7 agent environments (MCP, Search, Terminal, SWE, Web, OS, Android) within a single model. Environment modeling is the training o

We open-source Qwen-AgentWorld-35B-A3B (MoE, 35B/3B active, 256K context) and AgentWorldBench. Two routes, one roadmap: 🔬 Build the simulator — scalable, controllable, surpassing real environments 🧠 I

auto-research style proposal loops should be data driven! they largely work best only when Data/Evals/Feedback give a useful gradient to hill-climb against increasingly auto-research is a very good to

self-harness: harnesses that improve themselves spoiler alert, it's all loop engineering! 1. weakness mining - run agent and observe failures 2. propose improvements to the harness 3. confirm improvem

Single-shot generation still surfaces net-new kernels with no public reference: NeMo vocab-parallel log-probs, Hyena context parallelism, SAM 3 mask suppression. One GEMM + All-Gather kernel hit 87.9µ

Great report on LLM agent communication protocols. Communication is a huge bottleneck in multi-agent systems. (worth bookmarking) The report builds a five-dimensional taxonomy (counterparty, payload,

The largest LLM-as-a-Judge reliability audit yet. Researchers ran 21 judges from nine providers over roughly 541,000 judgments on MT-Bench, JudgeBench, and RewardBench. Findings: Validating a judge wi

>> Scalable Evaluation for AI Agents << If you run agent evaluation in production, this one is worth your time. It shows that front-loading human judgment into reusable evaluation assets is useful. Bu

🚨🚨🚨A research project idea! How to measure world models? Everyone's talking about world models these days. World model here, world model there. We can argue about what 'world model' actually means, an

// Evolving Meta-Skill for Multi-Agent Systems // Can a multi-agent system get better at orchestration without touching a single weight? Automatic MAS generation has been stuck between two bad options

GLM-5.2 is great at design (Opus level IMO). I am also starting to see great results with long-running tasks, too. How is this possible? I think there are a few clever hacks. But I just came across th

GLM 5.2 quantized GGUFs: https://huggingface.co/unsloth/GLM-5.2-GGUF GLM 5.2 on Open Router: https://openrouter.ai/z-ai/glm-5.2 Minion coding agent: https://github.com/Sentdex/minion A decent consolid

Over the last two weeks, both the U.S. Government and Anthropic took significant actions that demonstrated their power to control access to AI by restricting what others can do with frontier models. T

arXiv:2305.06145v2 Announce Type: replace Abstract: In cloth-changing person re-identification (CCReID), it is critical to learn clothes-invariant feature, which can provide discriminative ID features

GPU MODE has powered much of the public GPU kernel work online, with a permissive license from day one and generous credit from researchers, NVIDIA, AMD, and others. Today we’re moving our datasets to

// Self-Harness: Harnesses That Improve Themselves // (bookmark this one) Most of the agent scaffolds we rely on today are built once and remain frozen or mostly unchanged. The harness, like the skill

Kimi Code, our open-source coding agent, just got a major upgrade! 🔹One-line CLI install, zero setup, fast startup​ 🔹Drag in videos as coding context: reference-to-LUT, long-video-to-short, screen-rec

No. Not by itself. Sergey Brin is absolutely wrong. Transformers by themselves are not “sufficient” for AGI. Nobody uses transformers on their own anymore. Everybody is supplementing them tools and ha

The Matrix idea of keeping humans as batteries is obviously weird... we would be more useful as dice. LLMs default to very similar kinds of arguments & structure, and even different LLMs seem to colla

One of the more interesting takes on positive alignment that have recently come out-it’s long and interesting, combining philosophy and training setups (eg reward proposals), and worth a read. What ha

// Continual Learning Bench // One of the research areas with lots of investments is continual learning. While there are many efforts, there is very little progress in measuring it. So the big questio

https://x.com/CAIS/status/2060031683420999844?s=20 AI systems may soon help run economies, infrastructure, and military operations. But these systems are not reliably loyal or secure. An adversary can

New research from Renmin University. Treat skill selection as a harness in its own right. If you design skill routing for personal or edge agents, this work argues that the selection layer is a first-

// Agents' Last Exam // Agents' Last Exam is a living benchmark of over 1,000 economically valuable tasks, built with 250+ industry experts and mapped to the U.S. federal occupational taxonomy. The ha

Critical context on the new Anthropic blog: 1, AGI is *harder* than RSI (as used below). AGI: machine can do anything human can do, autonomously [not achieved] RSI (as used below): AI is a useful codi

I always appreciate the opportunity to discuss @LawZero_ and our approach to honest, reliable AI. Working on the Scientist AI with my brilliant colleagues at LawZero has made me very confident that we

Leaving aside the question of consciousness, the Ted Chiang piece has a reasonable point about moral atrophy if you let AI make choices. But it is also interesting in light of the fact that repeated r

We're presenting ParseBench at CVPR 2026! ParseBench is the most comprehensive document understanding benchmark for VLMs. ✅ It contains 2k pages of real-world enterprise documents ✅ It has comprehensi

New research from Google. Just shows the impressive results you can get from custom agent harnesses. LEAP wraps a general-purpose LLM in an agentic scaffold that grounds every step in the Lean compile

@saranormous codex is agi man https://youtu.be/cFNI2FORAc0 oneshotted this, no notes https://x.com/TheTuringPost/status/2061901518522188251?s=20 Love that in this high-tech age of AI @swyx still uses

// Scaling Behavior of Single LLM-Driven Multi-Agent Systems // Does adding more agents actually make a multi-agent system better? It's possible that collective intelligence emerges from interaction d

// State-Externalizing Harnesses // A new paradigm is emerging on how to effectively build agents and harnesses. If there is a state that the environment can maintain reliably, it probably doesn't bel

// Reusable Context Engineering // Context bloat quietly kills long-horizon runs, but you can fix it from the outside without fine-tuning the underlying agent. (bookmark this) Context management is us

Very good advice on self-improving agents. (bookmark it) This is something I am seeing in my own experiments with coding agents and harnesses for long-horizon tasks. What I have found is that stronger

// Scaling Laws for Agent Harnesses // If you build agent harnesses, this one is worth your time. (bookmark it) Most harness tuning treats every token and tool call as if volume is all that counts. Ne

AI systems may soon help run economies, infrastructure, and military operations. But these systems are not reliably loyal or secure. An adversary can make an AI work against its own operator. In our n

Our team at @AIatMeta is excited to announce ATLAS: one of the largest automated formalization efforts to date. ATLAS contains Lean 4 formalizations of both statements and proofs from 25+ mathematics

Too many business leaders believe that AI says what it means. And it’s odd because we naturally attribute a high number of human traits to AI, and yet we refuse to believe it can have hidden intent? 3

For over a decade, we’ve accepted that end-to-end backprop is the only way to train deep networks. But holding the entire network in memory all at once is why AI training is hitting a resource wall. W

Get a state of the art protein structure prediction in just 9 lines of code. No MSAs needed. I'm so excited to show the world what we've been working on the for the past months!! I'm going to highligh