arXiv 论文：证伪驱动的模型无关智能体协议，输出纪律决定决策下限
Output Discipline Sets the Floor: A Falsification-Guided, Model-Agnostic Agent Protocol for Auditable Hypothesis-Driven Decision Making
用中医辨证基准证明了件事：很多模型翻车不是能力不够，是输出纪律不行。他们设计的证伪协议能把 Kimi K3 的方差从 0.218 压到 0.007，写智能体的值得看看。
论文在中医辨证基准 TCMEval-SDT（200 个公开案例）上发现，两年模型缩放对零样本得分影响仅 -0.03 到 +0.08，输出纪律问题（推理不终止、格式不合规、方差大）在前沿模型中依然存在。作者提出模型无关的五阶段闭环框架，用候选×证据矩阵证伪协议约束：每个候选都要对照每条证据，至少一半候选被证伪才能得出结论。该协议在 DeepSeek-V4-Flash 上提升 0.267/0.271，把 Kimi K3 的运行间标准差从 0.218 降到 0.007，但对已有纪律的 Qwen3-4B/32B 造成 -0.087 到 -0.120 的损失。协议迁移到误信息数学推理时，以 89.8% 召回率、6.8% 误报率检出注入的假前提（MCC 0.54）。
Output Discipline Sets the Floor: A Falsification-Guided, Model-Agnostic Agent Protocol for Auditable Hypothesis-Driven Decision Making
What binds large language models on auditable, hypothesis-driven decision tasks, where every candidate can be checked against the evidence? On the public TCMEval-SDT benchmark for Traditional Chinese Medicine syndrome differentiation (200 cases), the binding constraint is output discipline and reasoning structure, not model capability alone: two years of model scaling move zero-shot scores by only -0.03 to +0.08, while output-discipline failures (non-terminating reasoning, format non-compliance, large run-to-run spread) persist even in frontier models. We test this with a model-agnostic agent framework formalizing decision making as a five-stage closed loop, enforced by a candidate x evidence matrix falsification protocol: every candidate is judged against every evidence item, at least half of the candidates must be falsified before a conclusion is licensed, and survivors are emitted in a machine-verifiable single-line format. Across five base models, the effect is sign-conditional on zero-shot output discipline: it rescues models whose discipline fails (DeepSeek-V4-Flash: +0.267/+0.271, entirely termination rescue), reduces a high-variance model's run-to-run standard deviation from 0.218 to 0.007 (Kimi K3), and significantly costs already-disciplined models (-0.087 to -0.120 on budget-matched Qwen3-4B/32B). Ablations attribute the cost to the candidate-bounded format layer and show the effect is independent of injected knowledge content (protocol-knowledge decoupling). The protocol also transfers to misinformation-robust mathematical reasoning, detecting injected false premises at 89.8% recall with a 6.8% false-positive rate (MCC 0.54; a flag-everything baseline scores 0) and restoring accuracy to the no-misinformation ceiling. Where zero-shot output discipline is weak, the reasoning protocol sets the floor; base-model capability always sets the ceiling.
- lmarena.ai10-12 18:24原文
