Free Question Answering on BARQA contextual
12.3BLEU ScoreGCoT-decoding
Evaluation results
| Method | Links | ||
|---|---|---|---|
| GCoT-decodingModel=Llama-3.1-8B, Decoding strategy=Multi-path2026.04 | 12.3 | 44.1 | |
| GCoT-decoding + SpanAlignModel=Llama-3.1-8B, Decoding strategy=Multi-path2026.04 | 10.9 | 41.5 | |
| GCoT-decoding + SpanAlignModel=Qwen2.5-14B, Decoding strategy=Multi-path2026.04 | 10.9 | 43.3 | |
| GCoT-decodingModel=Gemma-7B, Decoding strategy=Multi-path2026.04 | 10.9 | 37.7 | |
| GreedyModel=Llama-3.1-8B, Decoding strategy=Single-path2026.04 | 10.8 | 39.7 | |
| GreedyModel=Qwen2.5-14B, Decoding strategy=Single-path2026.04 | 10.7 | 44.4 | |
| GCoT-decodingModel=Qwen2.5-14B, Decoding strategy=Multi-path2026.04 | 10.2 | 38.9 | |
| Beam SearchModel=Qwen2.5-14B, Decoding strategy=Multi-path2026.04 | 10.1 | 42.1 | |
| Beam SearchModel=Llama-3.1-8B, Decoding strategy=Multi-path2026.04 | 10 | 38.5 | |
| Temperature samplingModel=Qwen2.5-14B, Decoding strategy=Single-path2026.04 | 7.7 | 42.5 | |
| Temperature samplingModel=Llama-3.1-8B, Decoding strategy=Single-path2026.04 | 7.3 | 37.4 | |
| Top-k samplingModel=Llama-3.1-8B, Decoding strategy=Single-path2026.04 | 6.8 | 37.2 | |
| Top-k samplingModel=Qwen2.5-14B, Decoding strategy=Single-path2026.04 | 6.4 | 40 | |
| GCoT-decoding + SpanAlignModel=Gemma-7B, Decoding strategy=Multi-path2026.04 | 5.8 | 36.5 | |
| GreedyModel=Gemma-7B, Decoding strategy=Single-path2026.04 | 4.7 | 36.6 | |
| Temperature samplingModel=Gemma-7B, Decoding strategy=Single-path2026.04 | 4.5 | 32.1 | |
| Beam SearchModel=Gemma-7B, Decoding strategy=Multi-path2026.04 | 4.2 | 35.4 | |
| Self-consistency + Prompt-basedModel=Llama-3.1-8B, Decoding strategy=Multi-path2026.04 | 3.6 | 30.4 | |
| Top-k samplingModel=Gemma-7B, Decoding strategy=Single-path2026.04 | 2.9 | 33.3 | |
| CoT-decoding + Prompt-basedModel=Llama-3.1-8B, Decoding strategy=Multi-path2026.04 | 2.4 | 25.1 | |
| Self-consistency + Prompt-basedModel=Gemma-7B, Decoding strategy=Multi-path2026.04 | 2.2 | 26.1 | |
| Self-consistency + Prompt-basedModel=Qwen2.5-14B, Decoding strategy=Multi-path2026.04 | 1.5 | 33.5 | |
| CoT-decoding + Prompt-basedModel=Qwen2.5-14B, Decoding strategy=Multi-path2026.04 | 1.4 | 32 | |
| CoT-decoding + Prompt-basedModel=Gemma-7B, Decoding strategy=Multi-path2026.04 | 0.7 | 21.5 |