Instruction Following on S-NI (SBERT Similarity)
60.2SBERT SimilarityDISTILLLENS
Evaluation Results
| Method | Links | |
|---|---|---|
| DISTILLLENSModel=Tiny-Llama-1.1B2026.02 | 60.2 | |
| JSDModel=Tiny-Llama-1.1B2026.02 | 57.8 | |
| AKLModel=Tiny-Llama-1.1B2026.02 | 57.7 | |
| SFTModel=Tiny-Llama-1.1B2026.02 | 57.6 | |
| KD (FKL)Model=Tiny-Llama-1.1B2026.02 | 56.8 | |
| RKLModel=Tiny-Llama-1.1B2026.02 | 56.8 | |
| FKL + RKLModel=Tiny-Llama-1.1B2026.02 | 55.8 | |
| DISTILLLENSModel=GPT-2 340M2026.02 | 55.1 | |
| SeqKDModel=Tiny-Llama-1.1B2026.02 | 54.7 | |
| SeqKDModel=GPT-2 340M2026.02 | 51.3 | |
| JSDModel=GPT-2 340M2026.02 | 50.4 | |
| FKL + RKLModel=GPT-2 340M2026.02 | 50.1 | |
| RKLModel=GPT-2 340M2026.02 | 49.8 | |
| AKLModel=GPT-2 340M2026.02 | 49.7 | |
| SFTModel=GPT-2 340M2026.02 | 49.4 | |
| KD (FKL)Model=GPT-2 340M2026.02 | 48.5 | |
| DISTILLLENSModel=GPT-2 120M2026.02 | 46.3 | |
| RKLModel=GPT-2 120M2026.02 | 44.2 | |
| FKL + RKLModel=GPT-2 120M2026.02 | 43.2 | |
| KD (FKL)Model=GPT-2 120M2026.02 | 40.6 | |
| JSDModel=GPT-2 120M2026.02 | 39.2 | |
| AKLModel=GPT-2 120M2026.02 | 39 | |
| SeqKDModel=GPT-2 120M2026.02 | 34.7 | |
| SFTModel=GPT-2 120M2026.02 | 32.9 |