arXiv cs.CL· T N Rickesh, Pranav Vinodh, Sowmya Kamath S·· 11 小时前AI 评分24
更少 Token，更多答案：面向多视角驾驶问答的 Ornith 风格 VLM 可验证奖励调优
Fewer Tokens, More Answers: Verifiable-Reward Tuning of an Ornith-Style VLM for Multi-View Driving QA
AI 导读
研究对适配多摄像头驾驶问答的 Ornith checkpoint 使用 GRPO 与可验证奖励进行调优，在 400 题开发集、256 token 上限下将答案提取率从 78.25% 提升至 90.25%，准确率从 55.50% 提升至 63.75%。
正文 · 原文
Abstract:Reasoning-enabled model derivatives must turn intermediate deliberation into a usable answer within a deployment budget. We study that transition in an Ornith checkpoint adapted to multi-camera driving question answering with verifiable rewards. Paired archived outputs show that, on a selected 400-question development screen with a 256-token cap, GRPO increases answer extraction from 78.25% to 90.25% and accuracy from 55.50% to 63.75%. Of 39 wrong-to-correct transitions, 33 begin with an unextractable base response. The base's 87 unextractable responses all reach the token cap. Checkpoint-verified reconstruction on those same questions gives a much smaller first-pass accuracy gain at 512 tokens: +0.75 percentage points, versus +8.25 at 256. A later answer-cued continuation is analyzed separately; it helps the base more and narrows the adapter difference. On the full 2,511-question development set, this recovery-enabled pipeline reduces mean generated tokens from 108.6 to 47.1, while the paired accuracy interval includes zero gain. We frame these findings as adaptation of answer delivery under finite budgets, with explicit selection and provenance limits. Ornith's self-scaffolding lineage and newer reasoning-enabled derivatives such as Bonsai motivate the problem, but our measured results concern Ornith alone and do not establish a shared architectural mechanism or cross-model transfer.
| Comments: | 7 pages, 2 figures, 3 tables |
| Subjects: | Computer Vision and Pattern Recognition (cs.CV); Computation and Language (cs.CL); Machine Learning (cs.LG) |
| Cite as: | arXiv:2611.00833 [cs.CV] |
| (or arXiv:2611.00833v2 [cs.CV] for this version) | |
| https://doi.org/10.48550/arXiv.2611.00833 arXiv-issued DOI via DataCite |
Submission history
From: Pranav Vinodh [view email]
[v1]
Sat, 31 Oct 2026 11:33:32 UTC (52 KB)
[v2]
Tue, 3 Nov 2026 03:56:47 UTC (52 KB)
来源：arXiv cs.CL · arxiv.org