清华与微软提出 HarnessMM，让智能体跳过重复 LLM 调用
清华和微软这篇论文教你的智能体少问 LLM：提前排好命令加检查条件，DeepSeek Harness 省了近 88% token。
清华与微软的论文提出 HarnessMM，让模型提前写好带精确参数和通过条件的后续命令，工具结果通过检查就直接执行，不再每次询问 LLM。在 98 个 HumanEval 任务上，DeepSeek Harness 成功率从 71.43% 升到 85.71%，token 消耗比原版少 87.93%。Codex 应用该方法后保持 90.82% 成功率，token 减少 29.54%。适合 write、build、test 这类可预测的固定执行链。
New Tsinghua and Microsoft paper finds that DeepSeek Harness used 87.93% fewer tokens than its original version on 98 HumanEval tasks while success rose from 71.43% to 85.71%.
Agent tools like Codex and DeepSeek Harness call the LLM after each tool result to choose the next step. Often that step was already decided, like running tests once the code compiles.
HarnessMM lets the model write future commands with exact arguments, plus a check for each, such as a successful build. When a tool result passes the check, the agent runs the command and skips the LLM call.
On 98 HumanEval tasks, Codex kept its 90.82% success while using 29.54% fewer tokens than plain Codex. DeepSeek Harness rose from 71.43% to 85.71% success and used 87.93% fewer tokens than the original.
If your agent runs predictable chains like write, build, test, have the model plan them once with pass checks instead of re-asking each step.
