Expired80% confidenceFactTime unknown
GPT-OSS 20B gave an incorrect answer in a classic logic reasoning test (determining who is lying).
1
Sources
80%
Confidence
Medium-term (~90 days)
Relevance
6/1/2026
First Seen
Valid until: 8/30/2026(expired)
Sources
Related Entities
Related Claims
UnverifiedIn a classic syllogism logic reasoning test, GPT-OSS gave the correct answer '不一定' (not necessarily) along with a counterexample explanation.80% similarUnverifiedIn a logical reasoning trap question test, GPT-5.5 produced an incorrect solution and failed to identify that the question had no answer.76% similarUnverifiedGPT-5.5 Pro produced a logically consistent proof of Erdős problem No. 710 within 22 minutes, but was in fact restating an old conclusion that had already been proven decades ago.70% similarUnverifiedIn a Mermaid flowchart generation test, both GPT-OSS 120B and 20B generated code with syntax errors, and the test failed.70% similarUnverifiedGPT-OSS 120B gave the correct answer to the classic water merchant reasoning problem, demonstrating a clear chain-of-thought reasoning process.70% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/23268API
curl https://kongchang.com/api/v1/knowledge/claims/23268MCP
get_claim(id=23268)