Unverified50% confidenceBenchmarkExact time
The two variants together achieve a significant advantage over the best unsupervised aggregator on 4 of the 6 benchmarks according to the Wilcoxon signed-rank test with p<0.05: DL19 is significant under momentum blending, while DL20, DBpedia, and NFCorpus are significant under full accumulated-history refit.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
10/30/2026
First Seen
Valid until: 1/28/2027
Sources
Related Entities
Related Claims
UnverifiedAfter testing 30 layer combinations, the best pairing was SenseNova L32 mapped to MiniMax H3 L49, using a 4096→128→5120 projector that achieved a cosine similarity of 0.9042 on the validation set.63% similarVerifiedFP8 has two variants: E4M3 (4-bit exponent + 3-bit mantissa), which has higher precision and is suited for forward inference, and E5M2 (5-bit exponent + 2-bit mantissa), which has a larger dynamic range and is suited for gradient computation.63% similarVerifiedBF16 uses 1 sign bit, 8 exponent bits, and 7 mantissa bits, sacrificing precision while retaining the same dynamic range as FP32.60% similarUnverifiedThe top five in MMLU-Pro are separated by only 0.6 percentage points: Darwin-180B-RSI 88.1%, MiniMax-M2.1 88.0%, Intern-S2-Preview 88.0%, Qwen3.5-397B-A17B 87.8%, DeepSeek-V4-Pro 87.5%.60% similarUnverifiedThe four features commonly used in 2048 solvers typically include the number of empty cells, monotonicity, smoothness, and the position of the maximum value.59% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1114250API
curl https://kongchang.com/api/v1/knowledge/claims/1114250MCP
get_claim(id=1114250)