Few-shot Language Understanding Suite (HS, WG, OBQA, ARC-E, ARC-C, PIQA, SIQA, Story) benchmark leaderboard · SOTA2 Research

Few-shot Language Understanding Suite (HS, WG, OBQA, ARC-E, ARC-C, PIQA, SIQA, Story)

0.555HS AccuracyTAD

Evaluation Results

Method	Links
TAD2026.02		0.555	0.602	0.352	0.636	0.326	0.714	0.449	0.694	0.541
TinyLlama(TL)-1.1B2026.02		0.552	0.589	0.334	0.613	0.307	0.714	0.421	0.689	0.527
TAD2026.02		0.551	0.6	0.348	0.642	0.331	0.713	0.448	0.691	0.54
TAD2026.02		0.548	0.599	0.338	0.625	0.32	0.721	0.442	0.687	0.535
TAD2026.02		0.546	0.6	0.346	0.632	0.324	0.718	0.446	0.689	0.538
Vanilla KD2026.02		0.541	0.589	0.335	0.632	0.316	0.712	0.442	0.687	0.532
CLM (no KD)2026.02		0.538	0.571	0.326	0.618	0.304	0.711	0.414	0.685	0.521
TinyLlama(TL)-1.1B2026.02		0.536	0.568	0.322	0.612	0.301	0.708	0.412	0.681	0.518