arXiv：cs.CL· Muyao Kong, Zhihao Lin·· 16 小时前
Scenario-Actor：让 LLM 先读懂场景再决定采样温度
Read the Room Before Your Model Generates
AI 导读
研究者提出可解释的温度选择框架 Scenario-Actor，从 Transformer 隐藏状态中提取场景结构，用场景嵌入向量确定解码温度。在 Llama-3.1-8B-Instruct 上，该方法在交互式社交场景目标完成、开放式指令跟随上取得显著提升，多轮对话表现与多个基线相当。
正文
Abstract:Large language models (LLMs) generally rely on sampling temperature to regulate the amount of randomness in their outputs. Mainstream methods for selecting temperature typically involve fixed, heuristic global settings and context or token-level calibration, rather than a geometrically interpretable representation of the communicative scenario. In this work, we introduce a simple and interpretable temperature selection framework, the Scenario-Actor, which adapts decoding to the inferred situational characteristics. Scenario-Actor extracts scenario-relevant structures from Transformer hidden states and determines the temperature from scenario embedding, a structured geometric representation of the communicative setting. Empirical results show that on Llama-3.1-8B-Instruct, Scenario-Actor achieves strong goal-completion performance in interactive social scenarios, significant gains on open-ended instruction following, and competitive performance on multi-turn conversation compared to multiple baselines.
| Subjects: | Artificial Intelligence (cs.AI); Computation and Language (cs.CL) |
| Cite as: | arXiv:2610.18820 [cs.AI] |
| (or arXiv:2610.18820v2 [cs.AI] for this version) | |
| https://doi.org/10.48550/arXiv.2610.18820 arXiv-issued DOI via DataCite |
Submission history
From: Muyao Kong [view email]
[v1]
Tue, 13 Oct 2026 17:51:18 UTC (693 KB)
[v2]
Wed, 14 Oct 2026 05:19:11 UTC (693 KB)
来源：arXiv：cs.CL · arxiv.org