跳到正文
研踪PaperTrail
← 研踪 PaperTrail

arXiv:2605.06196 · v2

The Granularity Axis: A Micro-to-Macro Latent Direction for Social Roles in Language Models

首次提交 2026-05-07 21:08
版本更新 2026-11-16 20:58
雷达观测 2026-11-17 17:11
arXiv ↗PDF ↗

摘要

Large language models (LLMs) are prompted to adopt social roles from individuals to institutions, but whether their representations encode this difference in social granularity remains unclear. We discover a contrast-based Granularity Axis, the difference between mean macro- and micro-role hidden states, that aligns with the dominant geometry of role space: in Qwen3-8B, its cosine with PC1 is 0.9720.972, and PC1 explains 52.6%52.6\% of the variance. Across 75 roles at five levels, with 90,000 role-conditioned and 1,200 default responses per model, projections increase monotonically and remain stable across layers, prompts, and response-quality filters. The structure transfers to Llama-3.1-8B-Instruct and scales to Qwen3-32B and Llama-3.1-70B-Instruct, whose PC1 shares are 45.99%45.99\% and 31.23%31.23\%. Independent five-rater validation (α=0.961α=0.961) preserves the ordering; identical continuations and affect/formality controls confirm scale-related variation beyond generated wording. The axis is also behaviorally causal: positive steering moves Llama from 2.002.00 to 3.173.17 on a five-point macro scale for prompts admitting local responses. Steering strength and stability depend on the model's default operating regime. Together, these findings establish social granularity as a representational primitive: a continuous, ordered, causally manipulable direction organizing role-conditioned generation across model families.

作者备注:37 pages; expanded experiments and analyses

AI 论文解读

基于摘要 · 非全文审阅

此版本尚未生成 AI 解读。解读由后台任务生成并按论文版本保存,登录后共享已有结果。

版本差异

本站首次收录为 v2,暂无更早版本可对照。

观测到的版本

相邻记录可能跳过中间版本;首次发现 v2 及以上也会标为版本更新。

v2 · 2026-11-16 20:58 · 最新观测版本

The Granularity Axis: A Micro-to-Macro Latent Direction for Social Roles in Language Models

Large language models (LLMs) are prompted to adopt social roles from individuals to institutions, but whether their representations encode this difference in social granularity remains unclear. We discover a contrast-based Granularity Axis, the difference between mean macro- and micro-role hidden states, that aligns with the dominant geometry of role space: in Qwen3-8B, its cosine with PC1 is 0.9720.972, and PC1 explains 52.6%52.6\% of the variance. Across 75 roles at five levels, with 90,000 role-conditioned and 1,200 default responses per model, projections increase monotonically and remain stable across layers, prompts, and response-quality filters. The structure transfers to Llama-3.1-8B-Instruct and scales to Qwen3-32B and Llama-3.1-70B-Instruct, whose PC1 shares are 45.99%45.99\% and 31.23%31.23\%. Independent five-rater validation (α=0.961α=0.961) preserves the ordering; identical continuations and affect/formality controls confirm scale-related variation beyond generated wording. The axis is also behaviorally causal: positive steering moves Llama from 2.002.00 to 3.173.17 on a five-point macro scale for prompts admitting local responses. Steering strength and stability depend on the model's default operating regime. Together, these findings establish social granularity as a representational primitive: a continuous, ordered, causally manipulable direction organizing role-conditioned generation across model families.

查看此版本解读 →查看此版本原文 ↗