Share your thoughts, 1 month free Claude Pro on usSee more

Commonsense Reasoning on ARC-e, ARC-c, OBQA, PIQA, HellaSwag (Zero-shot)

60.34Average Accuracy

LongChat-7B-v1.5

Updated 1mo ago

Evaluation Results

Method	Links
LongChat-7B-v1.5 2026.06		60.34
KIVI-2-gs32-r32 2026.06		60.3
STAR-KV (60%) 2026.06		60.12
STAR-KV (75%) 2026.06		59.18
KVQuant-2bit-1% 2026.06		58
Palu (50%) 2026.06		56.71
Palu (50%) 2026.06		56.42