Saved in:
| Main Authors: | , , |
|---|---|
| Format: | Preprint |
| Published: |
2025
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2509.24207 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866914426182959104 |
|---|---|
| author | Liu, Sijia Muennighoff, Niklas Ethayarajh, Kawin |
| author_facet | Liu, Sijia Muennighoff, Niklas Ethayarajh, Kawin |
| contents | Online alignment (e.g., GRPO) is generally more performant than offline alignment (e.g., DPO) -- but why? Drawing on prospect theory from behavioral economics, we propose a human-centric explanation. We prove that online on-policy sampling better approximates the human-perceived distribution of what the model can produce, and PPO/GRPO-style clipping -- originally introduced to just stabilize training -- recovers a perceptual bias in how humans perceive probability. In this sense, PPO/GRPO act as perceptual losses already. Our theory further suggests that the online/offline dichotomy is itself incidental to maximizing human utility, since we can achieve the same effect by selectively training on any data in a manner that mimics human perception, rather than restricting ourselves to online on-policy data. Doing so would allow us to post-train more quickly, cheaply, and flexibly without sacrificing performance. To this end, we propose a design pattern that explicitly incorporates perceptual distortions of probability into objectives like DPO/KTO/GRPO, creating humanline variants of them. Surprisingly, we find that these humanline variants, even when trained with offline off-policy data, can match the performance of their online counterparts (on both verifiable and unverifiable tasks) while running up to 6x faster. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2509_24207 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Humanline: Online Alignment as Perceptual Loss Liu, Sijia Muennighoff, Niklas Ethayarajh, Kawin Artificial Intelligence Online alignment (e.g., GRPO) is generally more performant than offline alignment (e.g., DPO) -- but why? Drawing on prospect theory from behavioral economics, we propose a human-centric explanation. We prove that online on-policy sampling better approximates the human-perceived distribution of what the model can produce, and PPO/GRPO-style clipping -- originally introduced to just stabilize training -- recovers a perceptual bias in how humans perceive probability. In this sense, PPO/GRPO act as perceptual losses already. Our theory further suggests that the online/offline dichotomy is itself incidental to maximizing human utility, since we can achieve the same effect by selectively training on any data in a manner that mimics human perception, rather than restricting ourselves to online on-policy data. Doing so would allow us to post-train more quickly, cheaply, and flexibly without sacrificing performance. To this end, we propose a design pattern that explicitly incorporates perceptual distortions of probability into objectives like DPO/KTO/GRPO, creating humanline variants of them. Surprisingly, we find that these humanline variants, even when trained with offline off-policy data, can match the performance of their online counterparts (on both verifiable and unverifiable tasks) while running up to 6x faster. |
| title | Humanline: Online Alignment as Perceptual Loss |
| topic | Artificial Intelligence |
| url | https://arxiv.org/abs/2509.24207 |