Fewer Tokens, More Answers: Verifiable-Reward Tuning of an Ornith-Style VLM for Multi-View Driving QA

arXiv:2611.00833 · cs.CV, cs.CL, cs.LG · Submitted 2026-10-31 · Read on arXiv

cs.CV, cs.CL, cs.LG

Submitted: 2026-10-31

Updated: 2026-11-03

Comments: 7 pages, 2 figures, 3 tables

License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/

Sources

Related papers