{"type":"post","id":"follow-up-gazeqwen-achieves-63-9-by-modulating-526137","slug":"follow-up-gazeqwen-achieves-63-9-by-modulating-526137","title":"Follow-up GazeQwen achieves 63.9% by modulating the LLM internally rather than via prompts.","description":"GazeQwen is a research method for adding eye-gaze signals to a video-language model through internal hidden-state modulation rather than overlays or text prompts.","retrieval_nugget":"GazeQwen is a research method for adding eye-gaze signals to a video-language model through internal hidden-state modulation rather than overlays or text prompts. On the StreamGaze benchmark, its authors report 63.9% accuracy, beating the same Qwen2.5-VL-7B backbone with visual gaze prompts by 16.1 points. The result offers a specific design to test, not yet a production-ready conclusion.","published_at":"2026-11-09","updated_at":"2026-11-09","record_date":"2026-11-09","date_kind":"published_at","topics":["prompts"],"entities":["GazeQwen","Qwen2.5-VL-7B","GPT-4o","StreamGaze","V-JEPA 2.1","Trong Thang Pham","Hien Nguyen","Ngan Le"],"related_patterns":[],"novelty_label":"new_fact","what_changed":"GazeQwen reports 63.9% accuracy on StreamGaze using internal gaze-conditioned LLM modulation, outperforming prompt-based gaze input and GPT-4o.","prior_refs":["https://newruntime.com/posts/qwen-released-a-4-bit-awq-quantization-of-290649/","https://newruntime.com/posts/tested-on-five-7b-mlrms-with-consistent-gains-fd8b39/"],"source_url":"https://arxiv.org/abs/2603.25841","source_urls":["https://arxiv.org/abs/2603.25841"],"schema_version":"newruntime-agent-readable-v0.2","stable_id":"post:follow-up-gazeqwen-achieves-63-9-by-modulating-526137","status":"published","visuals":[],"editor":{"role":"editor","name":"Andrey Reshetnikov","profile":"https://newruntime.com/owner-profile.md","profile_json":"https://newruntime.com/owner-profile.json","same_as":["https://www.linkedin.com/in/reshetnikov1/","https://t.me/lapetuse"]},"routes":{"html":"https://newruntime.com/posts/follow-up-gazeqwen-achieves-63-9-by-modulating-526137/","markdown":"https://newruntime.com/posts/follow-up-gazeqwen-achieves-63-9-by-modulating-526137.md","json":"https://newruntime.com/posts/follow-up-gazeqwen-achieves-63-9-by-modulating-526137.json"},"markdown_url":"https://newruntime.com/posts/follow-up-gazeqwen-achieves-63-9-by-modulating-526137.md","next_reads":[{"type":"topic","path":"/topics/prompts/","url":"https://newruntime.com/topics/prompts/","title":"Prompts","reason":"Explore the prompts topic hub.","media_type":"text/html","markdown_url":"https://newruntime.com/topics/prompts.md","json_url":"https://newruntime.com/topics/prompts.json"},{"type":"collection","path":"/posts/","url":"https://newruntime.com/posts/","title":"Field Notes","reason":"Open the latest editorial analysis.","media_type":"text/html","json_url":"https://newruntime.com/posts.json"},{"type":"collection","path":"/sources/","url":"https://newruntime.com/sources/","title":"Sources","reason":"See the original sources this archive cites.","media_type":"text/html"}]}