选择之前:研究语言模型中前瞻性表征的证据框架
Before the Choice: An Evidential Framework for Studying Prospective Representations in Language Models
Sion Israel Sion
中文摘要
可观察的决策揭示了所选动作,但并未确定内部表征了哪些备选方案。借鉴前瞻研究和海马前向扫描观察,我们开发了“选择之前”这一以证据为导向的框架,区分未来导向内容、前瞻性计算和因果引导。该框架组织了四个实验对比,并明确了其替代解释。首个测试案例涉及确定性符号环境中未被选择的轨迹。一个预注册的低秩探针在密封测试集上评估一次,未达到其标准(配对准确率0.020,而文本和观察选择基线为0.195;293个可接受对)。探索性H1b暴露了排除重建问题:识别所选候选可以确定未选目标。在H1c中,匹配世界变化了未选延续,同时保持候选键逐位不变。在指定阳性对照通过的情况下,晚键条件下的解码仍为0.521(p=0.18;141个贡献家族)。这些观察并未识别前瞻模拟或确立前瞻性表征的缺失。它们说明了为什么可识别性、敏感性和依赖感知的零模型必须约束从可解码性得出的推断。PREM被提出作为一个组织框架,并未被这些结果验证。我们将其作为对实证AI福利研究的方法论贡献,该研究将需要这样的证据区分。代码和协议记录:https://github.com/besionisrael/before-choice
关键要点
- 01问题:可观察决策无法揭示内部表征的备选方案,需区分前瞻性表征的证据。
- 02方法:提出“选择之前”框架,组织四个实验对比,并明确替代解释。
- 03结果:预注册低秩探针未达标准(配对准确率0.020 vs 基线0.195),晚键解码0.521(p=0.18)。
- 04局限:观察未识别前瞻模拟或缺失,强调可识别性、敏感性和依赖感知零模型的约束。
解读
尚无解读。
原始英文摘要
arXiv:2611.25170v1 Announce Type: new Abstract: An observable decision reveals the selected action but does not determine which alternatives were internally represented. Drawing on prospection research and hippocampal forward-sweep observations, we develop Before the Choice, an evidence-oriented framework distinguishing future-directed content, prospective computation, and causal guidance. The framework organizes four experimental contrasts and makes their alternative explanations explicit. Its first test case concerns unchosen trajectories in deterministic symbolic environments. A preregistered low-rank probe evaluated once on a sealed TEST did not meet its criterion (paired accuracy 0.020 versus 0.195 for a text-and-observed-choice baseline; 293 admissible pairs). Exploratory H1b exposed a reconstruction-by-elimination problem: identifying the chosen candidate can determine the unchosen target. In H1c, matched worlds varied an unchosen continuation while keeping candidate keys bitwise invariant. With specified positive controls passed, decoding in the late-key condition remained at 0.521 (p = 0.18; 141 contributing families). These observations do not identify prospective simulation or establish an absence of prospective representations. They illustrate why identifiability, sensitivity, and a dependence-aware null model must constrain inferences from decodability. PREM is proposed as an organizational framework and is not validated by these results. We offer it as a methodological contribution to empirical AI welfare research, which will require such evidential distinctions. Code and protocol records: https://github.com/besionisrael/before-choice