arXiv cs.CL· Roberto I. Ono Filho·· 13 小时前AI 评分31
每隔数百 token 注入新主题可提升基础语言模型的意外感与关联性评分
Interrupting the Loop: Periodic Subject Changes Raise Judged Surprise and Connection in Base Language Models
AI 导读
研究拆解了一个受认知启发的生成循环，在三个基础模型上测试 24 种条件，发现主要效果来自每隔数百 token 注入新主题（中断），在抑制字面重复（习惯化）的文本流上，该操作使 LLM 评判的意外感提升 1.2 至 1.4 分、关联性提升 0.7 至 0.9 分。
正文 · 原文
Abstract:Where does the novelty a base language model produces with no task come from, and what can an LLM judge of a long stream actually see? We dismantle a cognitively inspired generation loop over 24 conditions on three base models. Most of its effect lives in one operation: a new subject injected every few hundred tokens (an interruption) into a stream whose literal repetition is damped (habituation). We judge windows of generated text only, with the premise as the unit (n=10) and a judge measured for repeatability, against a second judge family and human readers. Under that protocol the interruption raises judged surprise by 1.2 to 1.4 points and connection by 0.7 to 0.9 over habituation alone. A connective that asks for continuity hurts; a bare paragraph break adds nothing detectable; a reset context does at least as well as a kept one; and a pre-registered replication on new premises confirms the primary contrast. Three things the window judge could not see changed the first version of this study. The judge scores the experimenter's injected sentence as the model's own. A fixed rotation of injected sentences makes the model replay its own earlier segments (65-80% of post-interruption windows at periods 150-300 are mostly verbatim replay), and the judge scores the replay as surprise and, above all, as connection; at period 300 the source of 62% of windows lies beyond the judge's 600-token horizon. And the local gains do not compose: no arm produces an integrated document. The salience monitor, the in-loop judge, memory across interruptions and a gated Review add nothing. On a problem with a verifier (online bin packing), the interruption multiplies valid, distinct candidate heuristics three- to fourfold without raising the quality of the best. We report an evaluation protocol for long generation and a characterization of a simple intervention, not a mechanism of creativity.
| Comments: | Version 2: corrections after an audit against the records (no conclusion changes; listed in the paper). 49 pages. Code: this https URL. Run data and LLM judgments: Zenodo https://doi.org/10.5281/zenodo.23163299%3B this version also at https://doi.org/10.5281/zenodo.23163693 |
| Subjects: | Computation and Language (cs.CL); Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2608.19893 [cs.CL] |
| (or arXiv:2608.19893v2 [cs.CL] for this version) | |
| https://doi.org/10.48550/arXiv.2608.19893 arXiv-issued DOI via DataCite |
Submission history
From: Roberto I Ono Filho [view email]
[v1]
Thu, 20 Aug 2026 11:01:59 UTC (2,528 KB)
[v2]
Thu, 22 Oct 2026 12:05:34 UTC (2,527 KB)
来源：arXiv cs.CL · arxiv.org