Generation-based extraction yields superior emotion vectors in small LMs that localize in middle layers following a U-shaped curve and enable architecture-dependent steering with three distinct behavioral regimes.
(2026b).MTI: A Behavior-Based Temperament Profiling System for AI Agents
2 Pith papers cite this work. Polarity classification is still indexing.
2
Pith papers citing it
fields
cs.CL 2years
2026 2verdicts
UNVERDICTED 2representative citing papers
Mature small language models share nearly identical 21-emotion geometries across architectures with Spearman correlations 0.74-0.92 despite opposite behavioral profiles, while immature models restructure under RLHF and prior comprehension-generation differences decompose into four distinct layers.
citing papers explorer
-
Extracting and Steering Emotion Representations in Small Language Models: A Methodological Comparison
Generation-based extraction yields superior emotion vectors in small LMs that localize in middle layers following a U-shaped curve and enable architecture-dependent steering with three distinct behavioral regimes.
-
Shared Emotion Geometry Across Small Language Models: A Cross-Architecture Study of Representation, Behavior, and Methodological Confounds
Mature small language models share nearly identical 21-emotion geometries across architectures with Spearman correlations 0.74-0.92 despite opposite behavioral profiles, while immature models restructure under RLHF and prior comprehension-generation differences decompose into four distinct layers.