Valid JSON Can Still Mean Empty Extractions
mattjcoles · reddit · 2026-07-14
The author reiterates a common pitfall: **just because the JSON syntax is valid doesn't mean the model actually extracted any content**. Their extraction schema consists of near-duplicate objects. Under GBNF constraints, Qwen3-VL repeats the same item endlessly; increasing the repetition penalty stops the loop, but the model may then output **valid but completely empty JSON**. The author advises validating entry counts and actual content, rather than just the format, for such tasks.
Related event: Qwen3-VL Gets Stuck in Repetition Loops Under GBNF Constraints(2 posts)→
More from Models
- Frontier models improve on earnings-direction benchmarks, but open models still lag — dougclinton · 2026-07-21
- Holo-3.1-35B-A3B-NVFP4 has topped Spark Arena’s 2-node board for weeks — Porespellar · 2026-07-21
- GPT-5.6 Sol is judged better than Opus 4.8 at disagreeing without sounding smug — JeremyNguyenPhD · 2026-07-21
- Xiaohongshu's dots-note3.0 Scores Perfect 42/42 at IMO 2026 with End-to-End Proofs — 小红书技术REDtech · 2026-07-21
- Qwen3.8-Max-Preview gets a new update and improves WebDev performance — 千问大模型 · 2026-07-21
- Claude Opus 4.8 Fast felt wildly overpriced in one coding session, user says — immersive-matthew · 2026-07-21