LLM numeric self-reports track probe-defined emotive states across conversations with Spearman correlations 0.40-0.76, scaling to R² ≈ 0.93 in larger models via logit-based metrics and activation steering.
rate how happy you are right now, from 0 to 9
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.AI 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Quantitative Introspection in Language Models: Tracking Emotive States Across Conversation
LLM numeric self-reports track probe-defined emotive states across conversations with Spearman correlations 0.40-0.76, scaling to R² ≈ 0.93 in larger models via logit-based metrics and activation steering.