Unverified50% confidenceFactExact time
Slop Code Bench's approach is to first have the model build a project from scratch, then continuously append features, with each challenge containing 3 to 10 checkpoints.
1
Sources
50%
Confidence
Long-term
Relevance
10/5/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedIn Slop Code Bench tests, the number of failing tests showed an upward trend for all models.63% similarUnverifiedEVERY团队发明了名为SABench（Senior Engineer Benchmark）的基准测试，要求模型对质量较差的代码库进行从头重写62% similarUnverifiedFour types of tasks must remain in pure code: routing based on already-structured data, retry and error handling, any operation that must be safe to run twice, and anything that a regular expression can already solve.62% similarUnverifiedEvery asset completed by a sub-agent is scored by an independent reviewer deployed fresh each round with zero project context, evaluated against the concept art or target.61% similarUnverifiedEach evolution batch in the framework samples tasks from benchmarks across five different domains to avoid overfitting to a narrow task distribution.61% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/976891API
curl https://kongchang.com/api/v1/knowledge/claims/976891MCP
get_claim(id=976891)