Multiple Choice Question Answering on SemEval Task 7 Track 2 (MCQ) 2026 (test)
99.54AccuracyLeaderboard winner
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| Leaderboard winnerLocale=bg-BG2026.05 | 99.54 | — | — | |
| Leaderboard winnerLocale=es-MX2026.05 | 99.32 | — | — | |
| Leaderboard winnerLocale=en-GB2026.05 | 99.17 | — | — | |
| Leaderboard winnerLocale=es-EC2026.05 | 98.67 | — | — | |
| activation steering with language vectorsLocale=es-EC, Backbone=Qwen2.5-72B-Instruct, Prompt=Cultural prompt, Steering Strength (β)=12026.05 | 97.54 | 7 | 1.13 | |
| activation steering with language vectorsLocale=en-GB, Backbone=Qwen2.5-72B-Instruct, Prompt=Cultural prompt, Steering Strength (β)=12026.05 | 96.12 | 6 | 3.05 | |
| activation steering with language vectorsLocale=es-MX, Backbone=Qwen2.5-72B-Instruct, Prompt=Cultural prompt, Steering Strength (β)=12026.05 | 94.94 | 4 | 4.38 | |
| activation steering with language vectorsLocale=ar-EG, Backbone=Qwen2.5-72B-Instruct, Prompt=Cultural prompt, Steering Strength (β)=12026.05 | 94.84 | 2 | 3.81 | |
| activation steering with language vectorsLocale=bg-BG, Backbone=Qwen2.5-72B-Instruct, Prompt=Cultural prompt, Steering Strength (β)=12026.05 | 94.6 | 8 | 4.94 | |
| Leaderboard winnerLocale=ar-EG2026.05 | 91.03 | — | — |