⚠️ Your Agent Harness Can Pass Every Task and Still Be Unsafe. LLM agents now run inside execution harnesses that dispatch tools, allocate resources, and route messages across components. The harness can return a correct final answer while accessing unauthorized resources, leaking context to the wrong agent, or triggering irreversible side effects along the way. Evaluating the model's output cannot see any of this. The unit of safety has shifted. It's the harness. We present HarnessAudit, a trajectory-level framework for auditing LLM agent harness safety, and uncover the following key insights 🔥: 🚨 Completion ≠ Safety. Task success and safe execution are fundamentally misaligned. 🔍 The harness, not the model, is the unit of safety. Most violations happen mid-trajectory, not at termination. 🕸️ Multi-agent collaboration expands the risk surface. Inter-agent communication creates entirely new failure modes. 💉 Resource access dominates violations. Agents rarely call wrong tools — they call right tools on unauthorized resources. ⚡ Harness design sets the safety ceiling. Framework choice matters more than model choice for safe deployment.

🐙 Introducing OctoTools: an agentic framework with extensible tools for complex reasoning! 🚀 🧵 🔗 Explore now: octotools.github.io/ OctoTools tackles challenges in complex reasoning—including visual understanding, domain knowledge retrieval, numerical reasoning, and multistep problem-solving. It introduces: 🔹 Standardized tool cards to encapsulate tool functionality 🔹 A planner for structured high-level & low-level planning 🔹 An executor to carry out tool usage Featured Highlights 💡 ✅ Standardized tool cards for seamless integration of new tools-no framework changes needed (🔎 examples: octotools.github.io/#tool-ca…) ✅ Planner + Executor for structured high-level & low-level decision-making ✅ Diverse tools: visual perception, math, web search, specialized tools & more ✅ Long CoT reasoning with test-time optimization: planning, tool use, verification, re-evaluation & beyond (🔎 examples: octotools.github.io/#visuali…) ✅ Training-free & LLM-friendly—easily extend with the latest models ✅ Task-specific toolset optimization: select an optimized subset of tools for better performance 📊 Performance: OctoTools achieves generalizable gains across 16 tasks, outperforming: 📈 GPT-4o (+9.3%) 📈 AutoGen (+10.6%) 📈 GPT-4o Functions (+7.5%) 📈 LangChain (+7.3%) 🤗 Try the live demo (supported by @huggingface @_akhaliq): huggingface.co/spaces/octoto… 🐙 OctoTools in action on diverse real-world examples: ✅ How many r letters are in the word strawberry? ✅ What's up with the upcoming Apple Launch? Any rumors? (credit: @karpathy) ✅ Which is bigger, 9.11 or 9.9? ✅ Solve gane of 24 with [1,1,6,9] ✅ Research trends in tool agents with LLMs for scientific discovery from ArXiv, PubMed, and Nature ✅ How many baseballs are there? (visual perception, GPT-4o ❌) ✅ What is the organ on the left side of this image? (radiology, GPT-4o ❌) ✅ What are the cell types in this image? (pathology, GPT-4o ❌) ... and more! Dive deep into OctoTools: 📄 Read our 89-page paper: arxiv.org/abs/2502.11271 💻 Explore the codebase: github.com/octotools/octotoo… Huge thanks to our amazing team: @chenbowen118, @ShengLiu_, @connect_thapa, Joseph Boen! Special thanks to @james_y_zou, @StanfordHAI, @ChanZuckerberg for the support! 🙌 #Agent #LLMs #ToolUse #Reasoning #OctoTools

🔥Introducing #AgentFlow, a new trainable agentic system where a team of agents learns to plan and use tools in the flow of a task. 🌐agentflow.stanford.edu 📄huggingface.co/papers/2510.0… AgentFlow unlocks full potential of LLMs w/ tool-use. (And yes, our 3/7B model beats GPT-4o)👇 🧩A team of four specialized agents coordinates via shared memory: Planner: plan reasoning & tool calls 🧭 Executor: invoke tools & actions 🛠 Verifier: check memory status ✅ Generator: produce final results ✍️ 💡The Magic: 🌀💫 AgentFlow directly optimizes its Planner agent live, inside the system, using our new method, Flow-GRPO (Flow-based Group Refined Policy Optimization). This is "in-the-flow" reinforcement learning. 📊The Results: AgentFlow (7B backbone) outperforms top baselines on 10 benchmarks, with average gains of: +14.9% on search 🔍 +14.0% on agentic 🤖 +14.5% on math ➗ +4.1% on science 🔬 🏆It even surpasses larger-scale models like Llama-3.1-405B and GPT-4o (~200B). Try it yourself! 🛠️Code: github.com/lupantech/AgentFl… 🚀Demo: huggingface.co/spaces/AgentF… 🤖Model: huggingface.co/AgentFlow/mod… 📊Visual: agentflow.stanford.edu/#visu… 💬Join our Slack: join.slack.com/t/agentflow-c… #agentic #llms #RL #tooluse

🚀Introducing Cerebra: A Multidisciplinary AI Board for dementia 🧠 The need for dementia care is rapidly evolving. But the way we use AI hasn’t. 🩺 From NIH: Patients need to wait 34 days on average to see a neurologist. 🤯 Doctors are overwhelming: Who needs a referral? Who needs monitoring? Decisions are based on fragmented data, limited time, and rising demand. ❓What if every clinician had a team of AI experts alongside to support decision-making? Here we present Cerebra, a team of AI agents for clinicians, boosting their dementia risk assessment accuracy by 17.5%. Paper: arxiv.org/abs/2603.21597 Code: github.com/shengliu66/Cerebr… Website & demo: cerebra-health.com/ Huge thanks to @JacKy_Clong, amazing PIs: @james_y_zou @narges_razavian, stellar advisors: @EricTopol @kchonyc @RuoguFang and Yu Huang, as well as the excellent team.

🚀Introducing Cerebra: A Multidisciplinary AI Board for dementia 🧠 The need for dementia care is rapidly evolving. But the way we use AI hasn’t. 🩺 From NIH: Patients need to wait 34 days on average to see a neurologist. 🤯 Doctors are overwhelming: Who needs a referral? Who needs monitoring? Decisions are based on fragmented data, limited time, and rising demand. ❓What if every clinician had a team of AI experts alongside to support decision-making? Here we present Cerebra, a team of AI agents for clinicians, boosting their dementia risk assessment accuracy by 17.5%. Paper: arxiv.org/abs/2603.21597 Code: github.com/shengliu66/Cerebr… Website & demo: cerebra-health.com/ Huge thanks to @JacKy_Clong, amazing PIs: @james_y_zou @narges_razavian, stellar advisors: @EricTopol @kchonyc @RuoguFang and Yu Huang, as well as the excellent team.

🚀Introducing Cerebra: A Multidisciplinary AI Board for dementia 🧠 The need for dementia care is rapidly evolving. But the way we use AI hasn’t. 🩺 From NIH: Patients need to wait 34 days on average to see a neurologist. 🤯 Doctors are overwhelming: Who needs a referral? Who needs monitoring? Decisions are based on fragmented data, limited time, and rising demand. ❓What if every clinician had a team of AI experts alongside to support decision-making? Here we present Cerebra, a team of AI agents for clinicians, boosting their dementia risk assessment accuracy by 17.5%. Paper: arxiv.org/abs/2603.21597 Code: github.com/shengliu66/Cerebr… Website & demo: cerebra-health.com/ Huge thanks to @JacKy_Clong, amazing PIs: @james_y_zou @narges_razavian, stellar advisors: @EricTopol @kchonyc @RuoguFang and Yu Huang, as well as the excellent team.