arXiv cs.CL· Jinkun Chen, Xuemin Yu, Sijia Han, Vlado Keselj·· 11 小时前AI 评分18
跨提示词复用实体替换响应：Qwen 与 Llama 上的隐藏状态迁移测量
Cross-Prompt Reuse of Entity-Replacement Responses
AI 导读
研究在模型权重固定下测量实体替换响应在英文事实性问题间的迁移量，将两个源问题的隐藏状态变化注入目标问题。在 0.5-1B Qwen 和 Llama 上，其他关系源隐藏复用分别消除 45% 和 49% 的无变化误差；Gemma-2-2B 消除 22%，恢复默认 BOS token 后达 58-68%。更大检查点、chat 格式和新实体上出现格式或队列相关失败。
正文 · 原文
Abstract:With model weights fixed, we measure how much of an entity-replacement response transfers between English factual questions, beyond whether answers change. We add the hidden-state change from two source questions to a target question about the same entity, comparing the predicted log-probability shift between two relations' published answers, given known answer prefixes, with the measured one. Hidden reuse injects the aligned change at 50% depth before the target's remaining layers; direct reuse adds source output-score changes. With other-relation sources, hidden reuse removes 45% and 49% of no-change error on 0.5-1B Qwen and Llama. Gemma-2-2B removes 22% and passes one of three comparisons in plain text without its default beginning-of-sequence (BOS) token. A later protocol restoring it yields 58-68%, passing every comparison on three cohorts. Larger checkpoints, chat format and new entities show replication and format- or cohort-dependent failures. On small plain-format checkpoints, other-relation hidden reuse mainly reproduces the shift both relations' questions share, without reliably lowering relation-selective error. With target-relation sources, hidden reuse's post hoc advantage over direct reuse is 0.2-0.9 nats outside chat format, unconfirmed and not replacement-specific. The same relation-neutral change at both questions' entity tokens lowers relation-selective error more than random or other-replacement changes, indicating receiving-question dependence.
| Comments: | 37 pages, including appendices |
| Subjects: | Computation and Language (cs.CL); Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2611.00352 [cs.CL] |
| (or arXiv:2611.00352v1 [cs.CL] for this version) | |
| https://doi.org/10.48550/arXiv.2611.00352 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Jinkun Chen [view email]
[v1]
Fri, 30 Oct 2026 20:32:53 UTC (331 KB)
来源：arXiv cs.CL · arxiv.org