Unverified50% confidenceFactExact time
In the round where MiniMax official version failed the inventory question, its actual actions were correct, but it added a code fence outside the final JSON, causing strict interface parsing to fail.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
10/20/2026
First Seen
Valid until: 1/18/2027
Sources
Related Entities
Related Claims
Unverified7B及更小的模型在工具调用场景下容易遗漏必填参数或输出不完整的JSON格式，导致Agent框架解析失败67% similarUnverifiedJSON mode and structured output only solve the problem of format validity; they do not guarantee semantic correctness and do not provide calibrated confidence, so they cannot support threshold policies.64% similarUnverifiedv1.5.0 fixes an issue where incomplete tool input JSON during streaming parsing of tool calls could remain in content blocks and cause parsing exceptions.64% similarUnverifiedOn the 7 prompts ending with "JSON only", Haiku 4.5 wrapped its answer in a Markdown code block every time, 154 out of 154; Haiku 5.5 returned bare JSON about 60% of the time and still had code fences the rest of the time.63% similarUnverifiedThe request body of the LLM Dev Utilities API's JSON repair endpoint has a nested structure: the outer layer is valid JSON, and the value of the inner text field is the broken JSON string to be repaired.63% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1077345API
curl https://kongchang.com/api/v1/knowledge/claims/1077345MCP
get_claim(id=1077345)