
Chengguang Gan, Zhixi Cai, Yunhao Liang, Hanjun Wei, QingHao Zhang, Shiwen Ni · arXiv (Cornell University) 2026 · 2026
DOI: 10.48550/arxiv.2611.16581
Counts differ because each database indexes a different set of publications. We treat OpenAlex as the canonical count; Google Scholar is not shown (no API, and crawling it violates its ToS).
Social scene datasets label people at two levels: what each person is doing and what the group is there for. Training on both levels at once is often assumed to help each of them. We test this assumption for vision-language models (VLMs) on JRDB-Social, treating person actions as the details and group intent as the whole. We fine-tune Qwen3.5 4B under eight training arms with three seeds, repeat the core arms on Qwen3.5 9B and Qwen3-VL 4B, and judge every arm with a verdict rule fixed before testing. Writing each member's actions before the intent raises intent macro-F1 from 30.1 to 31.0, and writing the intent first reaches 31.2, while action F1 stays within 0.2 points of an action-only model. Under the selected recipe, no arm trained with intent improves actions. The gain requires correct, image-grounded actions: empty action lists or actions borrowed from another group with the same intent lower intent F1. Forced-prefix tests, occlusion, member masking, linear probes and component grafting show that the intent follows the written actions, relies more on people and partly resides in the vision tower. Supplying actions as input instead of as targets hurts both levels.
arXiv (Cornell University) 2026 · 0 citations
arXiv (Cornell University) 2026 · 0 citations
No comments yet — start the discussion below.