Instruction Following on Dolly (SBERT Similarity)
71.4SBERT SimilarityDISTILLLENS
Evaluation results
| Method | Links | |
|---|---|---|
| DISTILLLENSModel=Tiny-Llama-1.1B2026.02 | 71.4 | |
| DISTILLLENSModel=GPT-2 340M2026.02 | 71.2 | |
| KD (FKL)Model=GPT-2 340M2026.02 | 70 | |
| FKL + RKLModel=Tiny-Llama-1.1B2026.02 | 70 | |
| AKLModel=Tiny-Llama-1.1B2026.02 | 70 | |
| RKLModel=Tiny-Llama-1.1B2026.02 | 69.8 | |
| SFTModel=GPT-2 340M2026.02 | 69.6 | |
| KD (FKL)Model=Tiny-Llama-1.1B2026.02 | 69.6 | |
| JSDModel=Tiny-Llama-1.1B2026.02 | 69.6 | |
| FKL + RKLModel=GPT-2 340M2026.02 | 69.5 | |
| SeqKDModel=GPT-2 340M2026.02 | 69.4 | |
| RKLModel=GPT-2 340M2026.02 | 69.4 | |
| JSDModel=GPT-2 340M2026.02 | 69.4 | |
| SFTModel=Tiny-Llama-1.1B2026.02 | 69.4 | |
| SeqKDModel=Tiny-Llama-1.1B2026.02 | 69.3 | |
| DISTILLLENSModel=GPT-2 120M2026.02 | 69.1 | |
| AKLModel=GPT-2 340M2026.02 | 69 | |
| RKLModel=GPT-2 120M2026.02 | 66.4 | |
| FKL + RKLModel=GPT-2 120M2026.02 | 66.1 | |
| AKLModel=GPT-2 120M2026.02 | 66 | |
| JSDModel=GPT-2 120M2026.02 | 65.9 | |
| KD (FKL)Model=GPT-2 120M2026.02 | 65.2 | |
| SeqKDModel=GPT-2 120M2026.02 | 64.7 | |
| SFTModel=GPT-2 120M2026.02 | 64.6 |