arXiv cs.CL· Wei Gao (Shibei Qiyuan AI Technology), Cheng Tian (Shibei Qiyuan AI Technology), Duan Yang (Xi'an Jiaotong University), Jianqing Yin (Shibei Qiyuan AI Technology)·· 2 天前AI 评分22
可写性与路径：向寻址记忆注入知识的对照研究
Writability and the Pathway: A Controlled Study of Knowledge Injection into Addressed Memory
AI 导读
一项对照研究测试了向寻址条件记忆注入知识的可写性：在冻结 GPT-2 上新增模块的 2×2 实验中，自然问题目标字符串命中率在仅行更新下平均 5.3%，无锚定投影更新为 28.7%，加锚定后为 22.0%，但严格答案得分均为零。
正文 · 原文
Abstract:Addressed conditional memory provides a local surface for post-hoc knowledge injection, but a successful write need not transfer to a new question or preserve untargeted behavior. This study combines a controlled GPT-2 comparison with production-memory experiments on Qwen3.8-Flash-Next-FP8 and DeepSeek-V4.1-Flash. Newly initialized modules attached to frozen GPT-2 are tested in a 2 x 2 design: row-only versus row-plus-projection updates, each with or without fixed module-residual anchoring. All arms reproduce 50 trained completions in each of three sampling seeds. Natural-question target-string hits average 5.3% for both row-only arms, 28.7% for unanchored projection updates, and 22.0% with anchoring, while strict answer scores remain zero. Anchoring lowers measured drift; lexical hits do not establish entity-specific retrieval. In the archived Qwen experiments, gradient writes make the designated first answer token top-ranked for 132 independently reset targets, comprising 93 new promotions and 39 already top-ranked targets. A separate matched 24-target comparison gives 13 new promotions with Adam versus two with SPSA at equal iteration counts but unequal computational procedures. An 80-row case includes saved continuations and a replay record, while reformulations bound its scope. Six DeepSeek zero-order cases yield one new target-token promotion and one regression. These results support bounded writing feasibility and separate trained expression, question-form behavior, and preservation; first-token rankings do not establish complete-answer success or universal conditions for writable memory.
| Comments: | 22 pages, 2 figures |
| Subjects: | Computation and Language (cs.CL) |
| Cite as: | arXiv:2610.21553 [cs.CL] |
| (or arXiv:2610.21553v1 [cs.CL] for this version) | |
| https://doi.org/10.48550/arXiv.2610.21553 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Cheng Tian [view email]
[v1]
Thu, 15 Oct 2026 09:14:29 UTC (78 KB)
来源：arXiv cs.CL · arxiv.org