Expired50% confidenceFactExact time
The file in that tool that actually records measurement results contains only 30 rows of data, of which 25 rows are the same strategy against the same test set and 5 rows are stub-against-stub (placeholder against placeholder)
1
Sources
50%
Confidence
Short-term (~14 days)
Relevance
9/17/2026
First Seen
This claim has expired. It is kept for the record and no longer describes the current state.
Valid until: 10/1/2026(expired)
Sources
Related Entities
Related Claims
UnverifiedSingle-precision floating-point (float, 32-bit) has only 23 mantissa bits, and can precisely represent integers in the range of approximately ±2^24.66% similarUnverifiedThe model ultimately outputs a tensor of 201×18×4, where 4 represents 4 lane lines, 18 represents key rows, and 201 represents 200 grid positions plus 1 non-existence class.65% similarUnverifiedRegardless of the encoding scheme used, 4-bit precision can only distinguish 16 different numerical levels.63% similarUnverified根据Artificial Analysis数据，Fable完成单个任务约用33k token，Opus 5约用37k token63% similarUnverifiedCaveman's fixed benchmark test results from 54 runs show a 33.2% reduction in input tokens, with all 18/18 correctness checks passing.62% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/928073API
curl https://kongchang.com/api/v1/knowledge/claims/928073MCP
get_claim(id=928073)