Studies
Agent Research
Research on autonomous agents, tool use and multi-agent systems
FilterCompanies & models
-
Realistic MCP Faults Halve Agent Quality, and a Bigger Model Doesn’t Help
Injecting MCP tool failures mined from 1,774 server repos cut agents’ task-completion score by 60%. gpt-oss-120B was no more resilient than gpt-oss-20B; the agent framework mattered more.
By
Editor's picks
- Realistic MCP Faults Halve Agent Quality, and a Bigger Model Doesn’t Help
- Stale CLAUDE.md and AGENTS.md Files Override the Code, and One Sentence Fixes It
- Coding Agents Can Be Tricked Into Running Attacker Commands After Compacting Context
- Even Top Agents Finish Only ~30% of Real Startup Product Workflows
- No Single Agent Memory Architecture Wins Everywhere, Systems Study Finds
More stories
-
Coding Agents Can’t Build the Agent You Actually Want — Yet
AgentSWE’s 25 commission-style tasks: strongest builders average 19.1 on Creation and 31.6 on Editing. Optimization closes under 5% of the remaining gap; 58.1% of held-out Creation…
-
Coding Agents Reach the Right File but Still Miss the Lines That Matter
SWE-Explore grades line-level exploration on 848 issues across 203 repos and 10 languages. Agentic explorers beat classical retrieval, but line coverage — not file hits —…
-
AI Agents Almost Never Stop in Time When a Task Is Impossible
Across 28,000+ tasks, timely abstention recall stays under 40% for all models. Best web AbsRec@1 is 26.7%; a scaffold can lift Llama-3.3-70B from 26.7% to 57.4%.
-
Agent “Skills” Mostly Anchor Procedures, Not Inject Knowledge — and Retrieval Breaks Them
Skills beat workflow memory by 6.06 points, but 65.7% of their benefit is procedural anchoring. As skill pools grow from 5 to 100, actual-use precision falls…
-
LLMs Can Build Their Own Agent Harnesses — But Still Trail Human Ones on Code
HarnessDev: six creator models build execution infrastructure from a seed. Generated harnesses match or beat references on writing and ML experiments, but lag on code and…
-
Coding Agents Fix More Bugs When Shown Whole Files, Study Finds
A UIUC study of 300 real GitHub issues finds coding agents resolve more bugs when shown a few whole, verbatim files than with prompt compressors or…