List your company

General Capability Evaluation on ARC-C, HellaSwag, MMLU, GSM8K

54.27ARC-C AccuracyPre-trained
39.39843.25947.1250.981Jun 1, 2026

Evaluation results

MethodLinks
Pre-trainedModel=Qwen-3.5-9B2026.0654.2759.6772.6286.268.19
AlphaTokenModel=Qwen-3.5-9B2026.0653.1357.7171.283.1466.3
Token Cleaning ICML 2025Model=Qwen-3.5-9B2026.0652.7655.0268.9183.5465.06
STM NeurIPS 2025Model=Qwen-3.5-9B2026.0652.6557.4370.6382.8765.9
XTF ICLR 2026Model=Qwen-3.5-9B2026.0652.5657.6170.5183.5966.07
Pre-trainedModel=Gemma-3-4B2026.0651.4556.8659.63751.23
ssTOKEN ICLR 2026Model=Qwen-3.5-9B2026.0651.3955.1770.5282.7264.95
LoRA ICLR 2022Model=Qwen-3.5-9B2026.0651.2756.469.681.9764.81
AlphaTokenModel=Gemma-3-4B2026.0651.155.1258.2236.5850.26
Token Cleaning ICML 2025Model=Gemma-3-4B2026.0650.8555.8455.943348.91
LESS ICML 2024Model=Gemma-3-4B2026.0650.7756.4554.2631.9848.37
ssTOKEN ICLR 2026Model=Gemma-3-4B2026.0650.5854.3157.3435.249.36
STM NeurIPS 2025Model=Gemma-3-4B2026.0650.356.1957.336.2150
LESS ICML 2024Model=Qwen-3.5-9B2026.0650.2457.1169.6881.2364.57
Standard FTModel=Gemma-3-4B2026.0650.1752.6353.8724.1245.2
Standard FTModel=Qwen-3.5-9B2026.0649.2554.2166.2477.9261.91
LoRA ICLR 2022Model=Gemma-3-4B2026.0649.1955.8156.7131.4448.29
XTF ICLR 2026Model=Gemma-3-4B2026.0648.753.8856.3334.4448.34
Pre-trainedModel=Llama-3.2-3B2026.064355.8256.5427.4545.7
AlphaTokenModel=Llama-3.2-3B2026.0642.7655.4856.2227.445.47
LESS ICML 2024Model=Llama-3.2-3B2026.0642.6254.6253.5722.6143.36
STM NeurIPS 2025Model=Llama-3.2-3B2026.0642.0254.6155.326.2244.54
ssTOKEN ICLR 2026Model=Llama-3.2-3B2026.0641.9453.3753.125.3843.45
Token Cleaning ICML 2025Model=Llama-3.2-3B2026.0641.7752.253.942643.48
LoRA ICLR 2022Model=Llama-3.2-3B2026.0641.6253.3250.422141.59
XTF ICLR 2026Model=Llama-3.2-3B2026.0640.4752.0350.6922.5841.44
Standard FTModel=Llama-3.2-3B2026.0639.9752.4548.2320.5840.31