List your company

Common-sense Reasoning on lm-eval-harness 11 tasks

63.430-shot AccuracyMoE Transformer w/ dynamic conv. (head-wise)
45.07449.839554.60559.3705Jun 2, 2026

Evaluation results

MethodLinks
MoE Transformer w/ dynamic conv. (head-wise)Train Tokens=100B, Params=6.80B, Conv. Location=QKV, Zero-shot=true2026.0663.43
MoE Transformer w/ dynamic conv. (low-rank)Train Tokens=100B, Params=6.80B, Conv. Location=QKV, Zero-shot=true2026.0663.42
MoE Transformer w/ static conv.Train Tokens=100B, Params=6.77B, Conv. Location=QKV, Zero-shot=true2026.0662.97
MoE TransformerTrain Tokens=100B, Params=6.77B, Conv. Location=–, Zero-shot=true2026.0662.46
Transformer w/ dynamic conv. (low-rank)Train Tokens=100B, Params=1.88B, Conv. Location=all linear, Zero-shot=true2026.0660.7
Transformer w/ dynamic conv. (low-rank)Train Tokens=100B, Params=1.88B, Conv. Location=QKV, Zero-shot=true2026.0659.7
Transformer w/ static conv.Train Tokens=100B, Params=1.83B, Conv. Location=QKV, Zero-shot=true2026.0658.94
Transformer w/ dynamic conv. (head-wise)Train Tokens=100B, Params=1.88B, Conv. Location=QKV, Zero-shot=true2026.0658.46
TransformerTrain Tokens=100B, Params=1.82B, Conv. Location=–, Zero-shot=true2026.0658.35
Transformer w/ more params (wider FFN)Train Tokens=100B, Params=1.87B, Conv. Location=–, Zero-shot=true2026.0658.23
Gated DeltaNet w/ dynamic conv. (low-rank)Train Tokens=15B, Params=309.5M, Conv. Location=QKV, Zero-shot=true2026.0649.22
Transformer w/ dynamic conv. (low-rank)Train Tokens=15B, Params=311.8M, Conv. Location=QKV, Zero-shot=true2026.0648.9
Transformer w/ dynamic conv. (low-rank)Train Tokens=15B, Params=319.0M, Conv. Location=all linear, Zero-shot=true2026.0648.81
Transformer w/ dynamic conv. (head-wise)Train Tokens=15B, Params=311.7M, Conv. Location=QKV, Zero-shot=true2026.0647.61
Mamba-2 w/ dynamic conv. (low-rank)Train Tokens=15B, Params=309.8M, Conv. Location=QKV, Zero-shot=true2026.0647.34
Gated DeltaNet w/ dynamic conv. (head-wise)Train Tokens=15B, Params=309.6M, Conv. Location=QKV, Zero-shot=true2026.0647.27
TransformerTrain Tokens=15B, Params=305.2M, Conv. Location=–, Zero-shot=true2026.0647.26
Mamba-2 w/ dynamic conv. (head-wise)Train Tokens=15B, Params=309.8M, Conv. Location=QKV, Zero-shot=true2026.0647.24
Gated DeltaNet (w/o conv.)Train Tokens=15B, Params=305.2M, Conv. Location=–, Zero-shot=true2026.0646.93
Transformer w/ static conv.Train Tokens=15B, Params=305.4M, Conv. Location=QKV, Zero-shot=true2026.0646.86
Gated DeltaNet w/ static conv.Train Tokens=15B, Params=305.4M, Conv. Location=QKV, Zero-shot=true2026.0646.76
Transformer w/ more params (wider FFN)Train Tokens=15B, Params=311.5M, Conv. Location=–, Zero-shot=true2026.0646.64
Mamba-2 (w/o conv.)Train Tokens=15B, Params=306.2M, Conv. Location=–, Zero-shot=true2026.0646.41
Mamba-2 w/ static conv.Train Tokens=15B, Params=306.4M, Conv. Location=QKV, Zero-shot=true2026.0645.78