List your company

Summarization on XSum (ROUGE-L)

27.45ROUGE-LZO-Adam
12.68216.51620.3524.184May 5, 2026May 7, 2026May 10, 2026May 13, 2026May 16, 2026May 19, 2026May 22, 2026

Evaluation results

MethodLinks
ZO-AdamBackbone=Qwen-3, Model Size=8B, Fine-tuning=16-rank LoRA2026.0527.45
RAdaZOBackbone=Qwen-3, Model Size=8B, Fine-tuning=16-rank LoRA2026.0527.34
MEAZOBackbone=Qwen-3, Model Size=8B, Fine-tuning=16-rank LoRA2026.0527.31
ZO-SGDBackbone=Qwen-3, Model Size=8B, Fine-tuning=16-rank LoRA2026.0526.96
RAdaZOBackbone=Llama-3.2, Model Size=1B, Fine-tuning=16-rank LoRA2026.0522.39
MEAZOBackbone=Llama-3.2, Model Size=1B, Fine-tuning=16-rank LoRA2026.0522.39
ZO-AdamBackbone=Llama-3.2, Model Size=1B, Fine-tuning=16-rank LoRA2026.0522.36
ZO-SGDBackbone=Llama-3.2, Model Size=1B, Fine-tuning=16-rank LoRA2026.0521.99
RAdaZOBackbone=Qwen-3, Model Size=0.6B, Fine-tuning=16-rank LoRA2026.0520.12
ZO-SGDBackbone=Qwen-3, Model Size=0.6B, Fine-tuning=16-rank LoRA2026.0520.08
ZO-AdamBackbone=Qwen-3, Model Size=0.6B, Fine-tuning=16-rank LoRA2026.0520.08
MEAZOBackbone=Qwen-3, Model Size=0.6B, Fine-tuning=16-rank LoRA2026.0520.08
BaseBackbone=Qwen-3, Model Size=8B, Fine-tuning=None (Before tuning)2026.0518.6
ZO-SGDBackbone=Llama-3.2 1B, Tuning=20-token prompt tuning2026.0517.56
ZO-AdamBackbone=Llama-3.2 1B, Tuning=20-token prompt tuning2026.0517.14
RAdaZOBackbone=Llama-3.2 1B, Tuning=20-token prompt tuning2026.0516.9
MEAZOBackbone=Llama-3.2 1B, Tuning=20-token prompt tuning2026.0516.85
BitDelta-col+distillRank=r16, BPW=1.00, Size (MB)=2.9, Time=+15 min2026.0516.78
ZO-SGDBackbone=Qwen-3 0.6B, Tuning=20-token prompt tuning2026.0516
MEAZOBackbone=Qwen-3 0.6B, Tuning=20-token prompt tuning2026.0515.72
ZO-AdamBackbone=Qwen-3 0.6B, Tuning=20-token prompt tuning2026.0515.62
PTQ-LoRDBARank=r16, BPW=2.0, Size (MB)=5.8, Time=+2 min2026.0515.55
QAT FreezeRank=r16, BPW=1.5, Size (MB)=4.4, Time=+4h2026.0515.54
PTQ-LoRDBARank=r16, BPW=1.5, Size (MB)=4.4, Time=+2 min2026.0515.52
LoRAQuant ρ=0.9Rank=r16, BPW=1.71, Size (MB)=5.0, Time=+17 min2026.0515.49
QAT FullRank=r16, BPW=2.0, Size (MB)=5.8, Time=+7h2026.0515.49
LoRA FP16Rank=r16, BPW=16.0, Size (MB)=46.4, Time=–2026.0515.38
QAT FreezeRank=r16, BPW=4.0, Size (MB)=11.6, Time=+4h2026.0515.38
QAT FullRank=r16, BPW=4.0, Size (MB)=11.6, Time=+7h2026.0515.35
QAT FullRank=r16, BPW=1.5, Size (MB)=4.4, Time=+7h2026.0515.34
PTQ-LoRDBARank=r16, BPW=4.0, Size (MB)=11.6, Time=+2 min2026.0515.19
QAT FreezeRank=r16, BPW=2.0, Size (MB)=5.8, Time=+4h2026.0515.18
BaseBackbone=Qwen-3, Model Size=0.6B, Fine-tuning=None (Before tuning)2026.0515.09
BaseBackbone=Qwen-3 0.6B, Tuning=20-token prompt tuning2026.0515.09
QAT ScratchRank=r16, BPW=4.0, Size (MB)=11.6, Time=7h2026.0515.07
QAT ScratchRank=r16, BPW=2.0, Size (MB)=5.8, Time=7h2026.0515.05
LoRAQuant ρ=0.5Rank=r16, BPW=1.22, Size (MB)=3.6, Time=+8 min2026.0515.02
QAT ScratchRank=r16, BPW=1.5, Size (MB)=4.4, Time=7h2026.0514.72
QAT ScratchRank=r64, BPW=1.0, Size (MB)=11.6, Time=7h2026.0514.54
BaseBackbone=Llama-3.2, Model Size=1B, Fine-tuning=None (Before tuning)2026.0513.62
BaseBackbone=Llama-3.2 1B, Tuning=20-token prompt tuning2026.0513.62
RAdaZOBackbone=Qwen-3 0.6B, Tuning=20-token prompt tuning2026.0513.25