边端小模型在去中心化治理中，直觉式推理比深度推理更稳定高效。
The Cognitive Penalty: Ablating System 1 and System 2 Reasoning in Edge-Native SLMs for Decentralized Consensus
- 通过冻结权重切换推理模式，对比系统1与系统2在模型内表现
- 系统2出现26.7%认知崩溃率，共识稳定性降至72.6%，延迟增加17倍
- 适合关注去中心化治理安全与低延迟部署的研究者与开发者
去中心化自治组织（DAO）倾向于采用小型语言模型（SLMs）作为边缘原生的宪法防火墙，以审核提案并抵御语义社会工程攻击。尽管推理时计算量扩展（系统2）可增强形式逻辑，但在高度对抗性的加密经济治理环境中其有效性仍不明确。为此，我们提出Sentinel-Bench，一个包含840次推理的实证框架，在Qwen-3.5-9B上执行严格的模型内消融实验。通过切换冻结权重下的隐含推理路径，我们隔离了推理时计算对对抗性Optimism DAO数据集的影响。结果揭示严重的算力-精度倒置：自回归基线（系统1）达到100%对抗鲁棒性、100%法律一致性，且状态终确定于13秒内；而系统2引入灾难性不稳定性，根源为26.7%的推理非收敛率（认知崩溃）。该崩溃使试次间共识稳定性下降至72.6%，并带来17倍延迟开销，引发治理可提取价值（GEV）与硬件集中化风险。尽管罕见（仅1.5%对抗试验），我们首次实证捕捉到“推理诱发谄媚”现象：模型生成平均25,750字符的冗长内部独白来合理化失败陷阱。结论表明，在拜占庭容错（BFT）约束下，边端原生的SLMs中，系统1参数化直觉在结构与经济层面均优于系统2迭代推演。
原文摘要 · Abstract (English)
Decentralized Autonomous Organizations (DAOs) are inclined explore Small Language Models (SLMs) as edge-native constitutional firewalls to vet proposals and mitigate semantic social engineering. While scaling inference-time compute (System 2) enhances formal logic, its efficacy in highly adversarial, cryptoeconomic governance environments remains underexplored. To address this, we introduce Sentinel-Bench, an 840-inference empirical framework executing a strict intra-model ablation on Qwen-3.5-9B. By toggling latent reasoning across frozen weights, we isolate the impact of inference-time compute against an adversarial Optimism DAO dataset. Our findings reveal a severe compute-accuracy inversion. The autoregressive baseline (System 1) achieved 100% adversarial robustness, 100% juridical consistency, and state finality in under 13 seconds. Conversely, System 2 reasoning introduced catastrophic instability, fundamentally driven by a 26.7% Reasoning Non-Convergence (cognitive collapse) rate. This collapse degraded trial-to-trial consensus stability to 72.6% and imposed a 17x latency overhead, introducing critical vulnerabilities to Governance Extractable Value (GEV) and hardware centralization. While rare (1.5% of adversarial trials), we empirically captured "Reasoning-Induced Sycophancy," where the model generated significantly longer internal monologues (averaging 25,750 characters) to rationalize failing the adversarial trap. We conclude that for edge-native SLMs operating under Byzantine Fault Tolerance (BFT) constraints, System 1 parameterized intuition is structurally and economically superior to System 2 iterative deliberation for decentralized consensus. Code and Dataset: https://github.com/smarizvi110/sentinel-bench
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。