kingy tested
Try the correction
Make this JSON parse.
Start with an original failed response, or paste a non-sensitive example. This runs in your browser; your text is not sent to Kingy.
Rejects duplicate keys and non-finite numbers. It removes one whole-response fence; it cannot correct invented or wrong values.
What happened
Symptom
The parser rejects a response starting with ```json.
Conditions
Qwen3-0.6B-Q8_0 and Qwen3-1.7B-Q8_0 in the retained four-task JSON pilot
Correction
A bounded post-processing diagnostic removes one whole-response Markdown code fence, then reruns strict JSON parsing and exact-answer validation. This does not repair wrong values.
Result
Retained diagnostics show these six enclosing-fence cases matched the answer key once unfenced. Original strict scores remain unchanged.
What did not work
Three formatting-only failures for 0.6B and three for 1.7B remained failed in the original benchmark.
- Preserve the original response and its failed strict score.
- Only strip a single enclosing fence if it wraps the entire response.
- Parse with duplicate-key and nonstandard-number rejection.
- Compare keys, values, types and array order against the source answer; fail closed on any mismatch.
Sources & limits
Small Qwen3 models: retained JSON pilot
Record checked 2026-09-07.
This is post-processing verified against retained output, not a retested prompting correction.
Do not strip arbitrary text, silently accept wrong values, or use an unknown answer key as proof.