List your company

Zero-shot text reconstruction on Chinese official news text n=200 chunks (test)

99.1BERTScore F1Entropy-in-FreqBuckets
55.440866.775478.1189.4446May 27, 2026

Evaluation results

MethodLinks
Entropy-in-FreqBucketsRetention rate (rkeep)=0.9, Decoder=Gemini 2.0 Flash2026.0599.1
StepRetention rate (rkeep)=0.9, Decoder=Gemini 2.0 Flash2026.0598.43
Word LengthRetention rate (rkeep)=0.9, Decoder=Gemini 2.0 Flash2026.0598.36
Word FrequencyRetention rate (rkeep)=0.9, Decoder=Gemini 2.0 Flash2026.0598.31
Qwen SFTRetention rate (rkeep)=0.9, Decoder=Qwen SFT2026.0598.28
Entropy-LPRetention rate (rkeep)=0.9, Decoder=Gemini 2.0 Flash2026.0598.23
EntropyRetention rate (rkeep)=0.9, Decoder=Gemini 2.0 Flash2026.0598.18
Opt (w/ polish)Retention rate (rkeep)=0.9, Decoder=Gemini 2.0 Flash2026.0598.03
Opt (w/o polish)Retention rate (rkeep)=0.9, Decoder=Gemini 2.0 Flash2026.0597.67
Hybrid-0.3Retention rate (rkeep)=0.9, Decoder=Gemini 2.0 Flash2026.0597.6
Hybrid-0.7Retention rate (rkeep)=0.9, Decoder=Gemini 2.0 Flash2026.0597.47
Hybrid-0.5Retention rate (rkeep)=0.9, Decoder=Gemini 2.0 Flash2026.0597.42
Word FrequencyRetention rate (rkeep)=0.7, Decoder=Gemini 2.0 Flash2026.0595.14
Word LengthRetention rate (rkeep)=0.7, Decoder=Gemini 2.0 Flash2026.0595.02
Qwen baseRetention rate (rkeep)=0.9, Decoder=Qwen base2026.0595.02
StepRetention rate (rkeep)=0.7, Decoder=Gemini 2.0 Flash2026.0594.68
Entropy-LPRetention rate (rkeep)=0.7, Decoder=Gemini 2.0 Flash2026.0593.95
Opt (w/ polish)Retention rate (rkeep)=0.7, Decoder=Gemini 2.0 Flash2026.0593.82
EntropyRetention rate (rkeep)=0.7, Decoder=Gemini 2.0 Flash2026.0593.71
Opt (w/o polish)Retention rate (rkeep)=0.7, Decoder=Gemini 2.0 Flash2026.0593.6
Qwen SFTRetention rate (rkeep)=0.7, Decoder=Qwen SFT2026.0593.53
Entropy-in-FreqBucketsRetention rate (rkeep)=0.7, Decoder=Gemini 2.0 Flash2026.0592.35
Hybrid-0.3Retention rate (rkeep)=0.7, Decoder=Gemini 2.0 Flash2026.0592.18
Hybrid-0.7Retention rate (rkeep)=0.7, Decoder=Gemini 2.0 Flash2026.0590.87
Hybrid-0.5Retention rate (rkeep)=0.7, Decoder=Gemini 2.0 Flash2026.0590.68
Word FrequencyRetention rate (rkeep)=0.5, Decoder=Gemini 2.0 Flash2026.0587.03
Qwen SFTRetention rate (rkeep)=0.5, Decoder=Qwen SFT2026.0586.52
Word LengthRetention rate (rkeep)=0.5, Decoder=Gemini 2.0 Flash2026.0586.48
StepRetention rate (rkeep)=0.5, Decoder=Gemini 2.0 Flash2026.0586.36
Opt (w/ polish)Retention rate (rkeep)=0.5, Decoder=Gemini 2.0 Flash2026.0585.07
EntropyRetention rate (rkeep)=0.5, Decoder=Gemini 2.0 Flash2026.0584.98
Entropy-LPRetention rate (rkeep)=0.5, Decoder=Gemini 2.0 Flash2026.0584.96
Hybrid-0.3Retention rate (rkeep)=0.5, Decoder=Gemini 2.0 Flash2026.0584.81
Opt (w/o polish)Retention rate (rkeep)=0.5, Decoder=Gemini 2.0 Flash2026.0584.6
Qwen baseRetention rate (rkeep)=0.7, Decoder=Qwen base2026.0583.47
Hybrid-0.5Retention rate (rkeep)=0.5, Decoder=Gemini 2.0 Flash2026.0583.37
Entropy-in-FreqBucketsRetention rate (rkeep)=0.5, Decoder=Gemini 2.0 Flash2026.0582.53
Hybrid-0.7Retention rate (rkeep)=0.5, Decoder=Gemini 2.0 Flash2026.0581.91
Hybrid-0.3Retention rate (rkeep)=0.3, Decoder=Gemini 2.0 Flash2026.0576.59
Hybrid-0.5Retention rate (rkeep)=0.3, Decoder=Gemini 2.0 Flash2026.0576.21
Entropy-LPRetention rate (rkeep)=0.3, Decoder=Gemini 2.0 Flash2026.0575.94
Opt (w/ polish)Retention rate (rkeep)=0.3, Decoder=Gemini 2.0 Flash2026.0575.83
EntropyRetention rate (rkeep)=0.3, Decoder=Gemini 2.0 Flash2026.0575.59
Qwen SFTRetention rate (rkeep)=0.3, Decoder=Qwen SFT2026.0574.51
Opt (w/o polish)Retention rate (rkeep)=0.3, Decoder=Gemini 2.0 Flash2026.0574.36
Entropy-in-FreqBucketsRetention rate (rkeep)=0.3, Decoder=Gemini 2.0 Flash2026.0574.32
Hybrid-0.7Retention rate (rkeep)=0.3, Decoder=Gemini 2.0 Flash2026.0573.94
StepRetention rate (rkeep)=0.3, Decoder=Gemini 2.0 Flash2026.0573.89
Word FrequencyRetention rate (rkeep)=0.3, Decoder=Gemini 2.0 Flash2026.0573.84
Word LengthRetention rate (rkeep)=0.3, Decoder=Gemini 2.0 Flash2026.0573.51
Qwen baseRetention rate (rkeep)=0.3, Decoder=Qwen base2026.0568.31
Hybrid-0.3Retention rate (rkeep)=0.1, Decoder=Gemini 2.0 Flash2026.0566.99
Hybrid-0.5Retention rate (rkeep)=0.1, Decoder=Gemini 2.0 Flash2026.0566.65
Qwen baseRetention rate (rkeep)=0.5, Decoder=Qwen base2026.0566.32
Opt (w/ polish)Retention rate (rkeep)=0.1, Decoder=Gemini 2.0 Flash2026.0566.26
Qwen SFTRetention rate (rkeep)=0.1, Decoder=Qwen SFT2026.0566.1
Hybrid-0.7Retention rate (rkeep)=0.1, Decoder=Gemini 2.0 Flash2026.0565.94
Entropy-LPRetention rate (rkeep)=0.1, Decoder=Gemini 2.0 Flash2026.0565.23
StepRetention rate (rkeep)=0.1, Decoder=Gemini 2.0 Flash2026.0565.12
EntropyRetention rate (rkeep)=0.1, Decoder=Gemini 2.0 Flash2026.0564.55
Word LengthRetention rate (rkeep)=0.1, Decoder=Gemini 2.0 Flash2026.0563.71
Word FrequencyRetention rate (rkeep)=0.1, Decoder=Gemini 2.0 Flash2026.0563.68
Entropy-in-FreqBucketsRetention rate (rkeep)=0.1, Decoder=Gemini 2.0 Flash2026.0562.71
Opt (w/o polish)Retention rate (rkeep)=0.1, Decoder=Gemini 2.0 Flash2026.0558.17
Qwen baseRetention rate (rkeep)=0.1, Decoder=Qwen base2026.0557.12