Longitudinal Affect Assessment on SemEval-2026 Task 2 Subtask 1 (dev)
0.67Valence Correlation (r) Compositeemo 15 GPT-5
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| emo 15 GPT-5Approach=LLM-based user-agnostic, Backbone=GPT-52026.04 | 0.67 | 0.641 | 0.359 | 0.425 | |
| emo 20Approach=LLM-based user-aware, Backbone=GPT-OSS 120B2026.04 | 0.661 | 0.568 | 0.378 | 0.433 | |
| emo granular 15Approach=LLM-based user-agnostic, Backbone=GPT-OSS 120B2026.04 | 0.657 | 0.701 | 0.374 | 0.411 | |
| emo 15Approach=LLM-based user-aware, Backbone=GPT-OSS 120B2026.04 | 0.647 | 0.59 | 0.37 | 0.42 | |
| emo w/e split 15Approach=LLM-based user-agnostic, Backbone=GPT-OSS 120B2026.04 | 0.647 | 0.684 | 0.364 | 0.437 | |
| v/a joint 15Approach=LLM-based user-aware, Backbone=GPT-OSS 120B2026.04 | 0.643 | 0.572 | 0.292 | 0.495 | |
| v/a split w/e split 15Approach=LLM-based user-aware, Backbone=GPT-OSS 120B2026.04 | 0.642 | 0.648 | 0.364 | 0.426 | |
| v/a split 15Approach=LLM-based user-aware, Backbone=GPT-OSS 120B2026.04 | 0.64 | 0.617 | 0.368 | 0.431 | |
| emo dynamic 15Approach=LLM-based user-aware, Backbone=GPT-OSS 120B2026.04 | 0.637 | 0.635 | 0.365 | 0.437 | |
| emo 15Approach=LLM-based user-agnostic, Backbone=GPT-OSS 120B2026.04 | 0.629 | 0.685 | 0.351 | 0.417 | |
| emo dynamic 15Approach=LLM-based user-agnostic, Backbone=GPT-OSS 120B2026.04 | 0.629 | 0.713 | 0.356 | 0.435 | |
| emo 10Approach=LLM-based user-aware, Backbone=GPT-OSS 120B2026.04 | 0.617 | 0.655 | 0.38 | 0.431 | |
| emo w/e split 15Approach=LLM-based user-aware, Backbone=GPT-OSS 120B2026.04 | 0.616 | 0.642 | 0.334 | 0.421 | |
| Ising (expectation)Approach=MaxEnt2026.04 | 0.533 | 0.831 | 0.277 | 0.478 |