Tsinghua & Microsoft's HarnessMM cuts agent token use by up to 87.93% while raising success rate
rohanpaul_ai · x · 2026-11-02
The Tsinghua–Microsoft paper HarnessMM (arXiv:2610.32813) removes redundant LLM calls in agent harnesses: the model pre-writes future commands with exact arguments plus pass checks, so when tool results satisfy the check the agent executes directly without re-asking the LLM.
On 98 HumanEval tasks, Codex kept 90.82% success with 29.54% fewer tokens, while DeepSeek Harness rose from 71.43% to 85.71% success with 87.93% fewer tokens. Takeaway: for predictable write → build → test chains, plan once with pass checks instead of re-invoking the model each step.
More from coding & agent
- InfoQ: A Systemic Approach to Memory, Knowledge, and Context in Agentic AI Architectures — AdiPolak · 2026-11-02
- Autonomous's Harness Pro uses an ESP32-P4, not a mini PC, to monitor AI coding agents — dee_hw · 2026-11-02
- RL-trained LFM2.5-24B adapter matches accuracy with 20% fewer tool calls — yb2698 · 2026-11-02
- Debugging production AI agents: error surfaces at step 5, root cause at step 3 — Ok-Machine7983 · 2026-11-02
- RAGFlow 1.0 preview: from RAG engine to agent-native database — newpeak · 2026-11-02
- Kitsune: open-source MCP server that shares your Firefox with a coding agent — Sad-Interview9085 · 2026-11-02