Unverified50% confidenceFactExact time
One of the most common sources of LLMs outputting incomplete JSON is that the model has a maximum output length (max tokens), and once exceeded it is truncated, leaving brackets and quotes unclosed.
1
Sources
50%
Confidence
Long-term
Relevance
10/16/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedIn streaming LLM APIs, some relay stations silently insert data: [DONE] to end early, or stuff error messages such as {"error": "rate limit exceeded"} into JSON disguised as model answers.74% similarUnverifiedThe author summarises five common situations in production environments where LLMs return non-standard JSON: Markdown fences, polite prose prefixes, trailing commas or smart quotes, truncation, and silent garbage that is structurally correct but missing fields.69% similarUnverifiedLLM API requests have a strict limit on the amount of data per call, and untruncated error logs may exceed this limit, rendering the conversation unusable63% similarUnverifiedQuantity assertions based on business semantics (such as validating the number of chapters) often catch truncation issues better than pure JSON format validation.63% similarUnverifiedOnce final JSON parsing fails, subsequent semantic checks are skipped entirely, which is the technical root cause of the 'task completed yet zero score' phenomenon.63% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1045662API
curl https://kongchang.com/api/v1/knowledge/claims/1045662MCP
get_claim(id=1045662)