ABSA-mix: leaderboard
Aspect-based sentiment analysis test data (ABSA-mix; Liskowski and Jankowski, Snowflake, 2026) drawn from 17 public datasets across 92 domains plus synthetic mixed and unknown-sentiment examples. Given a text and a list of aspects, the model labels each aspect positive, negative, neutral, mixed or unknown and gives the text's overall sentiment as JSON; the score is accuracy in percent, higher is better. General LLMs are prompted zero-shot at temperature 0, and the Arctic-ABSA models are the authors' fine-tunes on the ABSA-mix training data.
Metric: Accuracy (%; five-class aspect and overall sentiment on the ABSA-mix test data). Source: arxiv.org. Saturation forecast: Estimated already saturated. 9 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude 3.5 Sonnet | 83.65 | #747 |
| 2 | Mistral Large 2 (Jul) | 82.77 | #1153 |
| 3 | GPT-4o | 82.15 | #741 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=absa-mix · How It Works · Data refreshed daily, snapshot 2026-11-12.