UKP_Psycontrol at SemEval-2026 Task 2: Modeling Valence and Arousal Dynamics from Text

TL;DR

This paper introduces a multi-method system for modeling affective states and their short-term changes from user-generated texts, combining LLM prompting, structured probabilistic models, and neural regression, achieving top rankings in SemEval-2026.

Contribution

It presents a novel combination of three approaches for affect modeling, demonstrating the effectiveness of recent affective trajectories over textual semantics in short-term affect prediction.

Findings

LLMs effectively capture static affective signals.

Short-term affect variation is better explained by recent numeric trajectories.

Our system ranked first in both Subtask 1 and Subtask 2A.

Abstract

This paper presents our system developed for SemEval-2026 Task 2. The task requires modeling both current affect and short-term affective change in chronologically ordered user-generated texts. We explore three complementary approaches: (1) LLM prompting under user-aware and user-agnostic settings, (2) a pairwise Maximum Entropy (MaxEnt) model with Ising-style interactions for structured transition modeling, and (3) a lightweight neural regression model incorporating recent affective trajectories and trainable user embeddings. Our findings indicate that LLMs effectively capture static affective signals from text, whereas short-term affective variation in this dataset is more strongly explained by recent numeric state trajectories than by textual semantics. Our system ranked first among participating teams in both Subtask 1 and Subtask 2A based on the official evaluation metric.