检测路由翻转比知道是否修复它更容易：量化混合专家模型中由因果路由介导的损伤
Detecting a Route Flip Is Easier Than Knowing Whether to Fix It: Causal Route-Mediated Damage in Quantized Mixture-of-Experts
AI总结：
该研究针对量化混合专家模型，提出因果分析工具量化路由介导的损伤比例，发现路由器可检测路由翻转但无法区分其利弊，存在收益检测障碍，相关特性具跨模型性且受架构调节。
AI中文摘要：
Top-k混合专家（MoE）的路由是不连续的，因此一种受部署驱动的数值干扰——受保护的BF16门控读取模拟4位KV缓存量化——会推动令牌越过决策边界，翻转触发的专家。本文未提出新的缓解措施，而是提供了一套因果分析工具、实证发现和检测极限结果。该工具通过四次运行来量化路由介导的损伤比例（RMF），这是一种按机制分解的令牌级归因，预先注册的探针将发现扩展到三种架构。在OLMoE-1B-7B的4位KV（试点）设置下，约三分之一的损伤是路由介导的：RMF约为0.31（发现值0.31，95%置信区间[0.20, 0.41]；过程复现均值0.313±0.020；预先注册的重新执行结果为0.231）。可部署的路由器边际能检测到翻转发生（AUC为0.772），但无法区分有害翻转和有益翻转（准确率与随机猜测相当）：在测试的局部、推理可观测的路由器统计量中，未发现任何翻转损失符号的预测因子超过随机水平——这是一个经验性的收益检测障碍，限制了仅基于该特征族的选择性修复。带符号翻转的损耗和符号不可区分性具有跨模型性；干净参考补救措施的收益受架构调节；受控的同检查点标志交换将门控的归一化约定重新定义为损伤幅度的调节因子，而非路由可恢复性机制。真实的int4 KV内核产生的比例与伪量化剂量曲线兼容，但效力不足（95%置信区间[-0.111, 0.394]包含零）——排除了重大分歧，而非独立复现。假设、阈值和评估在测量前已预先注册，且报告了遗漏情况；预先注册的保留集读取在样本外复现了划分和接近抵消的损耗，而严格的不可能排除仅差一点未达成。
英文摘要：
Top-k Mixture-of-Experts (MoE) routing is discontinuous, so a deployment-motivated numerical disturbance -- simulated 4-bit KV-cache quantization read by a protected BF16 gate -- pushes tokens across decision boundaries and flips which experts fire. This paper proposes no new mitigation; it supplies a causal apparatus, empirical findings, and a detection-limit result. A four-run apparatus prices the route-mediated fraction (RMF) of quantization damage, a token-level attribution decomposes it by mechanism, and pre-registered probes carry the findings across three architectures. On OLMoE-1B-7B at 4-bit KV (pilot), about a third of the damage is routing-mediated: RMF ~ 0.31 (discovery 0.31 [0.20, 0.41]; process-replicated mean 0.313 +/- 0.020; pre-registered re-execution 0.231). The deployable router margin detects that a flip occurred (AUC 0.772) but cannot tell a harmful flip from a helpful one (at chance): among the tested local, inference-observable router statistics we find no predictor of a flip's loss sign above chance -- an empirical benefit-detection barrier bounding selective repair restricted to this feature family. The signed-flip tax and sign-inseparability carry cross-model; the clean-reference remedy's payout is architecture-modulated; a controlled same-checkpoint flag-swap re-scopes the gate's normalization convention to a damage-magnitude moderator, not a route-recoverability mechanism. A real int4 KV kernel yields a fraction compatible with the fake-quant dose curve but underpowered (95% CI [-0.111, 0.394] includes zero) -- ruling out gross disagreement, not an independent replication. Hypotheses, thresholds, and evaluations were pre-registered before measurement, with misses reported; a pre-registered held-out read replicates the partition and the near-cancelling tax out of sample, while the strict impossibility exclusion narrowly misses.
