Instruction Following on Vicuna (SBERT Similarity)
73.6SBERT SimilarityDISTILLLENS
Evaluation Results
| Method | Links | |
|---|---|---|
| DISTILLLENSModel=Tiny-Llama-1.1B2026.02 | 73.6 | |
| FKL + RKLModel=Tiny-Llama-1.1B2026.02 | 70.2 | |
| RKLModel=Tiny-Llama-1.1B2026.02 | 70.1 | |
| JSDModel=Tiny-Llama-1.1B2026.02 | 70.1 | |
| AKLModel=Tiny-Llama-1.1B2026.02 | 70.1 | |
| DISTILLLENSModel=GPT-2 340M2026.02 | 69.9 | |
| SFTModel=Tiny-Llama-1.1B2026.02 | 69 | |
| KD (FKL)Model=Tiny-Llama-1.1B2026.02 | 69 | |
| FKL + RKLModel=GPT-2 340M2026.02 | 67.8 | |
| SeqKDModel=Tiny-Llama-1.1B2026.02 | 67.2 | |
| JSDModel=GPT-2 340M2026.02 | 66.7 | |
| KD (FKL)Model=GPT-2 340M2026.02 | 66.2 | |
| SeqKDModel=GPT-2 340M2026.02 | 65.9 | |
| RKLModel=GPT-2 340M2026.02 | 65.9 | |
| DISTILLLENSModel=GPT-2 120M2026.02 | 65.8 | |
| AKLModel=GPT-2 340M2026.02 | 65.3 | |
| FKL + RKLModel=GPT-2 120M2026.02 | 62 | |
| AKLModel=GPT-2 120M2026.02 | 61.7 | |
| RKLModel=GPT-2 120M2026.02 | 61.6 | |
| JSDModel=GPT-2 120M2026.02 | 61.6 | |
| SFTModel=GPT-2 120M2026.02 | 59.8 | |
| KD (FKL)Model=GPT-2 120M2026.02 | 58.9 | |
| SeqKDModel=GPT-2 120M2026.02 | 58.1 | |
| SFTModel=GPT-2 340M2026.02 | 55 |