让语言模型通过隐藏时间信号做带时序的任务决策
ChronoState: Hidden Elapsed-Time Conditioning for Temporal-State Action Selection in Frozen-Backbone Language Models
- 用隐藏时间变量与任务状态组合，让冻结模型做时序动作选择
- 隐藏时间注入使准确率达0.9305，显著高于无时间或乱序时间的对照组
- 适合研究时序推理、系统级控制的开发者，但不支持跨场景自动计时
语言模型在执行时序任务时，常需同时考虑符号化任务状态和已过去的时间（如缓存过期、配额重置、截止时间等）。本文研究是否可将经过的秒数tau作为非令牌的系统级标量，通过隐藏通道注入冻结的大型语言模型，并与可见符号状态融合。提出ChronoState基准：符号状态在提示中，时间以31维正弦+对数编码形式通过隐式时序注入通道输入，模型从预设动作中选择。采用Qwen2.5-3B-Instruct作为冻结的bf16骨干，结合门控FiLM残差调制和8秩LoRA动作表面。隐藏时间注入达到0.9305±0.0134的准确率与0.9410±0.0103的平衡准确率；无时间与打乱时间对照组分别降至0.5511±0.0042与0.3323±0.0097，且打乱时间下错误一致性强，表明模型确实在训练分布内依赖注入标量。对未见模板、时长及多约束组合仍有较强泛化能力，但未见配额家族迁移性能弱，仅0.5065±0.0559；而公平对比的提示+LoRA时间基线达0.9893±0.0052。结论为：在直接监督下，隐藏时间可与符号状态组合，但无法实现自主计时、广泛抽象或优于提示嵌入的时间戳。
原文摘要 · Abstract (English)
Temporal decisions in language-model systems often depend on both symbolic task state and elapsed wall-clock time, such as cache expiration, job completion, quota resets, deadlines, or stale sessions. We study whether elapsed time can be supplied as a non-token, system-side scalar and composed with visible symbolic state by a frozen-backbone language model. We introduce ChronoState, a compositional temporal-state benchmark in which symbolic state appears in the prompt, elapsed seconds tau are supplied through a hidden chronometric-injection channel, and the model selects a forced-choice temporal action. Here, "hidden" means hidden from the user-visible token sequence, not from model computation. Using Qwen2.5-3B-Instruct as a frozen bf16 backbone with a 31-dimensional sinusoidal-plus-log time encoding, gated FiLM residual modulation, and a rank-8 LoRA action surface, hidden-time CI reaches 0.9305 +/- 0.0134 accuracy and 0.9410 +/- 0.0103 balanced accuracy. No-time and shuffled-time controls fall to 0.5511 +/- 0.0042 and 0.3323 +/- 0.0097, respectively, with high shuffled-time wrong-state consistency supporting causal dependence on the injected scalar within the trained distribution. Generalization remains strong for held-out templates, durations, and multi-constraint compositions, but held-out quota-family transfer is weak at 0.5065 +/- 0.0559, while a fair prompt+LoRA timestamp baseline reaches 0.9893 +/- 0.0052. Thus, ChronoState supports a narrow conclusion: hidden elapsed time can be composed with symbolic task state under direct supervision, but does not establish autonomous time tracking, broad unseen-family abstraction, or superiority over prompt-injected timestamps.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。