How to stop LLMs from hallucinating values for fields missing from a report
Traditional_Push2492 · reddit · 2026-11-11
When extracting a long list of fields from 200-400 page reports (e.g. emissions figures), models fabricate numbers for missing fields instead of saying "not found." Extra passes are too costly, confidence scores are unreliable, and parsers like llamaparse/docling can drop tables on ugly pages, making blank ambiguous.
The proposed candidate fix: require a quote with each extracted value and verify it in code. This is a求助-style discussion thread.
More from coding & agent
- Burkov's new RL book implements REINFORCE, PPO and DQN on one rocket-landing task — burkov · 2026-11-11
- Gemini 4 Carbon rebuilds 220m of downtown Los Gatos in Blender with near-zero manual modeling — bilawalsidhu · 2026-11-11
- AgentRun goes open source as a managed agents platform supporting ACP harnesses — garrytan · 2026-11-11
- Fable 5.2 built an entire mobile game from scratch — playable now — brandon_galang · 2026-11-11
- Particle-1 cuts code review cost to 2-3 cents, ~40x cheaper than Greptile — garrytan · 2026-11-11
- Industry will regret not hiring junior developers, warns 3-5 year prediction — bendee983 · 2026-11-11