跳到正文
研踪PaperTrail
← 研踪 PaperTrail

arXiv:2609.36724 · v2

Routing in Gradient Space: Balanced Usage Is Not Expert Specialization

首次提交 2026-09-29 13:01
版本更新 2026-11-16 10:50
雷达观测 2026-11-17 17:12
arXiv ↗PDF ↗

摘要

Sparse expert models can distribute traffic evenly while still grouping incompatible training signals within the same experts. We study routing as a gradient-partitioning problem and introduce gradient-aligned routing (GAR), whose load-normalized router objective rewards grouping observations with aligned gradients. On five multi-task text-classification mixtures, we compare GAR with task-loss-only routing, gradient-combination and gradient-conflict methods, and load-balancing losses. With a fully trainable RoBERTa backbone and classification-head experts, GAR has the highest aggregate validation accuracy, 1.07 percentage points above task-loss-only routing. With frozen DeBERTa and Qwen3-1.7B backbones and low-rank adapter experts, it again ranks first, 1.10 points above task-loss-only routing, with better-balanced expert load and higher gradient-mass purity, the share of each expert's gradient-norm mass from its dominant task; the load-balancing losses flatten load further but leave this purity near its task-loss-only level. Top-1 routing, trainable full-parameter feed-forward network (FFN) experts, and a larger backbone also show positive aggregate gains. In component ablations, the objective's cross-observation alignment terms alone retain GAR's accuracy, and GAR keeps its accuracy when observations are formed without task identity; supervising the router with task labels raises task-expert association without adding accuracy. The results distinguish expert-load balance from gradient-based routing organization and show the predictive value of gradient-informed routing in multi-task text classification.

作者备注:58 pages (9 main, 49 appendix), 6 figures, 54 tables

AI 论文解读

基于摘要 · 非全文审阅

此版本尚未生成 AI 解读。解读由后台任务生成并按论文版本保存,登录后共享已有结果。

版本差异

本站首次收录为 v2,暂无更早版本可对照。

观测到的版本

相邻记录可能跳过中间版本;首次发现 v2 及以上也会标为版本更新。

v2 · 2026-11-16 10:50 · 最新观测版本

Routing in Gradient Space: Balanced Usage Is Not Expert Specialization

Sparse expert models can distribute traffic evenly while still grouping incompatible training signals within the same experts. We study routing as a gradient-partitioning problem and introduce gradient-aligned routing (GAR), whose load-normalized router objective rewards grouping observations with aligned gradients. On five multi-task text-classification mixtures, we compare GAR with task-loss-only routing, gradient-combination and gradient-conflict methods, and load-balancing losses. With a fully trainable RoBERTa backbone and classification-head experts, GAR has the highest aggregate validation accuracy, 1.07 percentage points above task-loss-only routing. With frozen DeBERTa and Qwen3-1.7B backbones and low-rank adapter experts, it again ranks first, 1.10 points above task-loss-only routing, with better-balanced expert load and higher gradient-mass purity, the share of each expert's gradient-norm mass from its dominant task; the load-balancing losses flatten load further but leave this purity near its task-loss-only level. Top-1 routing, trainable full-parameter feed-forward network (FFN) experts, and a larger backbone also show positive aggregate gains. In component ablations, the objective's cross-observation alignment terms alone retain GAR's accuracy, and GAR keeps its accuracy when observations are formed without task identity; supervising the router with task labels raises task-expert association without adding accuracy. The results distinguish expert-load balance from gradient-based routing organization and show the predictive value of gradient-informed routing in multi-task text classification.

查看此版本解读 →查看此版本原文 ↗