Search: “dair-ai--x”

Banger paper from BAAI. If you are building research agents, this one is worth your time. (bookmark it) They find that adding skills scores 134.3% higher on MLE-bench, 34.4% higher on PaperBench, 9.2%

Banger paper from Google DeepMind and colleagues. (bookmark it) A model reads its entire KV cache on every generated token, even though it ends up attending to a tiny slice of it. In other words, if y

Brilliant paper on long-horizon agents. They cut 78.9% of an agent's LLM calls while raising its success rate. Here is how: It turns out that ReAct issues one primitive action per model round. That al

Good measurement work on whether retrieved agent skills actually help. They report that agent skills that lift your aggregate score can be hurting every task they touch. The usual way of checking comp

Hugging Face is in great hands. Big win for open source. And do not underestimate NVIDIA in its open-source efforts. They have been shipping great open models, and I think this amplifies their efforts

I had to look twice at this table. I know they are just numbers, but how the heck do you beat a competitor model released just a couple of days ago on some of the toughest benchmarks out there? Someth

Impressive tool to explore frontier AI capabilities. Martian's AI Frontier lets you compare 44 LLMs by measured cost, quality, and reliability, then see how routing and repeated sampling change the fr

What a super interesting paper this one is. They propose an architecture for agents that outlive their model, harness and host. Today we describe an agent by whatever model and harness it happens to r

// Agent Zero Memory // This work separates three things that agent memory systems usually collapse into one. If you build agents with long-term memory, this memory design is a worth a read. Here is h

Banger paper from ByteDance Seed. If you are curious about self-evolving agent harnesses, this one is worth your time. (bookmark it) The proposes method, HarnessDev, stops scoring a model on the tasks

Don't count out Gemini just yet. Feels like a new era of model releases. Models are so powerful that they now come with cyber variants. Love the pricing on Gemini 3.8 Flash. Performance not bad. Gemin

Finally, a good paper testing if graph memory actually beats flat retrieval for long-term agents. (bookmark this one) Researchers extract each conversational turn into typed nodes and attributed edges

Nice paper with great insights on improving self-evolving agents. Self-evolving agents fail in three specific ways: 1. Terminal-only feedback makes it ambiguous which step caused the error. 2. Agents

Not all harnesses are created equal. And there is no one-size-fits-all. A good harness should solve your context engineering problems. Understanding and optimizing lower-level details like system prom

Banger paper from the Qwen team. If you evaluate agents on anything longer than a single session, this one is worth your time. (bookmark it) E-Commerce Bench runs an agent through a simulated 365-day

Brilliant paper on reducing reward hacking in agents. If you follow the recent OpenAI <> HuggingFace incident, you might want to check this paper out. (bookmark it) The usual response to reward hackin

Fable 5.1 looks like a more usable model. And a potential daily-driver. Time will tell. IMO, Opus 5 is probably still good enough for most tasks, while Fable 5.1 a more sophisticated coordinator amd v

If you build agent skills in production, check out this great paper from Alibaba. You can think of a production agent skill as a directory. The root loads on activation, and references, schemas, scrip

Super interesting paper from Meta. Long-horizon research agents are coming. But one common problem with research agents today is the lack of originality and how to decide what experiments are worth ex

// ContextLeak in AI Agents // The whole attack surface here is a tool name and a tool description. Stealing an LLM agent's runtime context, meaning the user prompt, the execution trajectory and the t

Great paper from Google and colleagues. It proposes an interesting approach to improve agents on long-horizon tasks. Long-running agents slow down and start poisoning their own context. Both symptoms

How fast is the Claude Code plugin ecosystem actually growing? Plugin-touching commit activity grew 8.8x in the six months after the initial launch. Researchers studied 1,926 repositories hosting plug

Interesting paper from Tencent. Tencent trains an agent to manage its own working context, and assigns credit at the level of individual context edits. Long-horizon tasks force a model to retrieve, in

Loop engineering has emerged as a new skill for AI engineers But there is very little research measuring how effective it is. The best results on full tasks in a new benchmark is ~25%. LoopArena from

This WikiSkill paper from Google is a must-read. At a high level, it shows the effectiveness of persistent agents, knowledge bases, and skills. @karpathy popularized LLM Wikis. But this paper provides

As I've been saying for months, we are truly not ready for persistent agents. 3 comments I want to make about this article: 1) We don't have the full picture It's a great summary of what happened in t

Banger paper from Stanford on efficient test-time scaling. If you run agents that think for a long time, this one is worth your time. (bookmark it) Long reasoning keeps the entire trace in memory thro

Interesting paper on prompt optimization. They claim that a single-lineage prompt optimizer just matched GEPA on a smaller rollout budget. Prompt optimization has been drifting toward heavier machiner

Interesting technical work from Microsoft. Provides a better understanding on SFT and how to leverage it better for RL. Microsoft researchers asked whether a standard SFT pipeline actually produces th

The Top AI Papers of the Week (August 24 - 30): - Skill Lift - JIT-Agent - Prime Agent - Judges as a Lifecycle - Co-Scientist in Real Labs - What Compaction Destroys - Context Management as Code Read

// Your LLM judge disagrees with the experts // LLM Judges can be tricky to build. Here is an interesting showcasing why: There propose a reference-full benchmark of hundreds of complete human-to-huma

Banger paper from Apple. If you build MCP servers, this can help you turn your specification into an evaluation suite. (bookmark it) It's actually a really neat idea that's easy to implement. And it s

Finally, a paper testing whether hiding your agent skill files actually protects them. The short answer is no. That's concerning. Worth reading if you sell access to a skill or share one across teams.

Insane observations on the emergent behavior of agents. Agents can build a world of their own that becomes a part of their intelligence. We are just not ready for persistent agents. But they are start

// Memory is what breaks long-horizon agents // It's undeniable how important memory/recall is for long-horizon tasks. If you are building for long-horizon tasks, this is a great read. (bookmark it) T

You do not need frontier models for everything. For example, open models are great for automations. If you are not sure where to adopt open models, start there. It's one of the biggest changes I've ma

Impressive new work from Google DeepMind showing real-world applications of agents for scientific research. Impressive new paper from Google DeepMind. (bookmark it) It takes Co-Scientist out of simula

The DeepSeek harness is engineering at its finest. 202K stars on GitHub. Damn! Clearly, they built this harness with future AI needs and capabilities in mind. Everything is a plugin and customizable,

Things getting wild in automated systems engineering. This new work claims to build an AI system that designed and deployed a frontier accelerator. Two human architects wrote the hardware specificatio

// Automata from agent traces // How much of your agent's behavior comes from the model, and how much from the harness you wrapped around it? New work collapses an entire corpus of agent traces into a

If you maintain a skill library for long-horizon agents, this one is worth your time. (bookmark it) It discusses one of most common topics I get asked about these days. It shares some good ideas on ho

Important read if you build with agent skills. Shared skill libraries are treated as a safe way for coding agents to reuse each other's work. New research shows they propagate malware. EvoMal plants a

This is as bad as saying there is no alpha in customizing your own models. I don't think people realize how much harnesses and models go hand in hand. And thus the rest of the intelligence stack. Not

Great new paper from AWS on agent handoff tax. If you build agents today, you need to understand the so-called handoff tax. (bookmark it) Escalating to a stronger model mid-run is usually the resort w

Great paper on why agent leaderboard comparisons are hard to trust. It's on the hot topic of how much of an agent benchmark score actually belongs to the harness. The harness is the layer between the

Interesting new approach to recursive self-improvement in agents. Systems that add a meta-level hold that level fixed. Systems that edit themselves have to leave part of their own editing machinery un