Instruction Following on UnNI (SBERT Similarity)
60.3SBERT SimilarityDISTILLLENS
Evaluation Results
| Method | Links | |
|---|---|---|
| DISTILLLENSModel=Tiny-Llama-1.1B2026.02 | 60.3 | |
| RKLModel=Tiny-Llama-1.1B2026.02 | 59.4 | |
| JSDModel=Tiny-Llama-1.1B2026.02 | 58.9 | |
| FKL + RKLModel=Tiny-Llama-1.1B2026.02 | 58.7 | |
| AKLModel=Tiny-Llama-1.1B2026.02 | 58.4 | |
| SFTModel=Tiny-Llama-1.1B2026.02 | 58.3 | |
| KD (FKL)Model=Tiny-Llama-1.1B2026.02 | 58 | |
| DISTILLLENSModel=GPT-2 340M2026.02 | 57 | |
| SeqKDModel=Tiny-Llama-1.1B2026.02 | 56.9 | |
| SeqKDModel=GPT-2 340M2026.02 | 54.4 | |
| RKLModel=GPT-2 340M2026.02 | 54 | |
| FKL + RKLModel=GPT-2 340M2026.02 | 54 | |
| JSDModel=GPT-2 340M2026.02 | 53.8 | |
| DISTILLLENSModel=GPT-2 120M2026.02 | 53.3 | |
| KD (FKL)Model=GPT-2 340M2026.02 | 53.3 | |
| AKLModel=GPT-2 340M2026.02 | 53.3 | |
| SFTModel=GPT-2 340M2026.02 | 53.1 | |
| FKL + RKLModel=GPT-2 120M2026.02 | 49.3 | |
| RKLModel=GPT-2 120M2026.02 | 49 | |
| KD (FKL)Model=GPT-2 120M2026.02 | 48.3 | |
| AKLModel=GPT-2 120M2026.02 | 48.2 | |
| JSDModel=GPT-2 120M2026.02 | 47.8 | |
| SFTModel=GPT-2 120M2026.02 | 45.9 | |
| SeqKDModel=GPT-2 120M2026.02 | 45.1 |