Unverified50% confidenceFactExact time
The digest function in the drift_check.py script uses SHA-256 and takes the first 16 hexadecimal characters (64 bits of entropy), which for distinguishing output text, a low-cardinality set, will almost never collide.
1
Sources
50%
Confidence
Long-term
Relevance
10/15/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedJanas supports speculative decoding: a small draft model guesses the next few tokens and the large model verifies them in one pass. On Qwen3-Next-80B this raises write speed from 24 to 37 tok/s, with the text content completely identical.59% similarUnverifiedcheckwash 0.6.0 achieved a precision of 0.67 and recall of 0.22 for its own block judgment on this corpus, while the raw count signal read independently yielded a precision of 0.80 and recall of 0.89.57% similarUnverifiedIn tests covering context lengths of 512–4096, increasing the draft length from 4 tokens to 16 tokens resulted in only a 0.2%–1.3% increase in overall inference latency.57% similarUnverifiedIn the task of completing the last Token of an FP8 quantization function, after DeepSeek-V4-Flash-Base adds equal-sign decorations before the code, when the number of equal signs modulo 4 is 0 or 1 the error rate is as high as 71.3%, and when the remainder is 2 or 3 the accuracy returns to 91.5%.56% similarUnverifiedThe question setter placed 256 (the result of 4^4 without adding 3) as a distractor option, differing from 259 by only 3.56% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1039887API
curl https://kongchang.com/api/v1/knowledge/claims/1039887MCP
get_claim(id=1039887)