arXiv 2610.29545 · cs.CL · 2026-10-21
PACE: Periodic and Channel-Aligned Encoding of Four-Bit Recurrent States
Patrick Haller, Filipe Laitenberger, Alan Akbik
Humboldt Universität zu Berlin
在最新榜 #366 本站收录于 2026-10-25
一句话要点
PACE 是线性注意力模型循环状态的训练后 4 比特量化：每隔几个 token 才取整一次并缓存其间的 token，再利用大值集中在少数列的规律按通道对齐编码，让同一加速器同时解码更多序列。
PACE is a post-hoc 4-bit quantizer for linear-attention recurrent states: it rounds only every few tokens with a small buffer in between and exploits column-clustered outliers via channel-aligned encoding, letting more sequences share an accelerator.
阅读 arXiv 原文 PDF Hugging Face 讨论
主题与学科
同主题：高效训练与推理
- The Teacher Is a Direction, Not a Destination: Extrapolating RL-Induced Representation Residuals in On-Policy Distillation ▲511
- Scaling Properties of Same-Family On-Policy Distillation ▲327
- Beyond Domain-Routed Supervision: Cross-Teacher Guarding for Multi-Teacher On-Policy Distillation ▲250
- On-Policy or Off-Policy Learning? A Systematic Study of Distillation Dynamics ▲205
- Beyond Teacher Assignment: Domain-Normalized Multi-Teacher On-Policy Distillation ▲181
同主题：大语言模型
- Jeva: System One Models from First Principles 2026-10-13 · ▲371
- Post-Training Leaves Behavioral Shadows on Unrelated Decisions 2026-09-24 · ▲273
- WeVisOne-VL: Unified Visual Perception from Objects to Pixels 2026-10-22 · ▲247
- Thinking Inertia: LLMs Keep Thinking When Told Not To 2026-10-08 · ▲241
- OmniEdu: Open Foundation Models for Learning and Teaching 2026-09-19 · ▲240
数据说明
元数据来自 arXiv API（已核实）；点赞来自 Hugging Face 每日论文；引用来自 Semantic Scholar。最近刷新 2026-10-26。
要点由本站根据摘要自写，不转载摘要原文，请阅读原文。