Saved in:
| Main Authors: | , , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2026
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2605.11601 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866910211281780736 |
|---|---|
| author | Lai, Wen Shen, Yingli Jin, Dingnan Cui, Qing Zhou, Jun Sun, Maosong Fraser, Alexander |
| author_facet | Lai, Wen Shen, Yingli Jin, Dingnan Cui, Qing Zhou, Jun Sun, Maosong Fraser, Alexander |
| contents | Autoregressive language models are widely used for text evaluation, however, their left-to-right factorization introduces positional bias, i.e., early tokens are scored with only leftward context, conflating architectural asymmetry with true text quality. We propose masked reconstruction as an alternative paradigm, where every token is scored using full bidirectional context. We introduce DiffScore, an evaluation framework built on Masked Large Diffusion Language Models. By measuring text recoverability across continuous masking rates, DiffScore eliminates positional bias and naturally establishes an evaluation hierarchy from local fluency to global coherence. We further provide diagnostic tools unavailable to autoregressive frameworks: multi-timestep quality profiles that decompose scores across masking rates, and bidirectional PMI decomposition that disentangles fluency from faithfulness. Experiments across ten benchmarks show that DiffScore consistently outperforms autoregressive baselines in both zero-shot and fine-tuned settings. The code is released at: https://github.com/wenlai-lavine/DiffScore. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2605_11601 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | DiffScore: Text Evaluation Beyond Autoregressive Likelihood Lai, Wen Shen, Yingli Jin, Dingnan Cui, Qing Zhou, Jun Sun, Maosong Fraser, Alexander Computation and Language Artificial Intelligence Autoregressive language models are widely used for text evaluation, however, their left-to-right factorization introduces positional bias, i.e., early tokens are scored with only leftward context, conflating architectural asymmetry with true text quality. We propose masked reconstruction as an alternative paradigm, where every token is scored using full bidirectional context. We introduce DiffScore, an evaluation framework built on Masked Large Diffusion Language Models. By measuring text recoverability across continuous masking rates, DiffScore eliminates positional bias and naturally establishes an evaluation hierarchy from local fluency to global coherence. We further provide diagnostic tools unavailable to autoregressive frameworks: multi-timestep quality profiles that decompose scores across masking rates, and bidirectional PMI decomposition that disentangles fluency from faithfulness. Experiments across ten benchmarks show that DiffScore consistently outperforms autoregressive baselines in both zero-shot and fine-tuned settings. The code is released at: https://github.com/wenlai-lavine/DiffScore. |
| title | DiffScore: Text Evaluation Beyond Autoregressive Likelihood |
| topic | Computation and Language Artificial Intelligence |
| url | https://arxiv.org/abs/2605.11601 |