Unverified50% confidenceBenchmarkExact time
Selecting for each task the agent design with the lowest perplexity, average perplexity increased by 3.4% over BF16, and downstream accuracy was 77.33%, compared with 79.56% for BF16.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
11/5/2026
First Seen
Valid until: 2/3/2027
Sources
Related Entities
Related Claims
UnverifiedIn Microsoft's internal evaluation covering 36 benchmarks and nearly 150,000 questions, Decision-1 ranked first with an average accuracy of 83.5%, surpassing Jev 1.13.0 (82.3%), Quyet-1.0-Large (81.9%), and GPT-6 Luna Decisions (79.4%).69% similarUnverifiedPerplexity Decider V1.1 27B topped the evaluation with a score of 80/82, a median latency of 0.445 seconds, and zero flips across 26 reversal tests.67% similarUnverifiedIn a write operation that scales feature encodings by three times, the graph built from previously active features has a median relative error of at least 0.247, while the maximum error of the margin graph is only 0.033.67% similarUnverifiedUnder an 18k character budget, anchored hybrid retrieval achieved 36.18% accuracy, BM25 achieved 35.98%, while the best direct planner-guided method reached only 34.19%.66% similarUnverifiedUnder BF16, if the result of any floating-point operation cannot be exactly represented in BF16, it must be rounded to the nearest representable value, producing a relative error on the order of about 2^-8.66% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1138469API
curl https://kongchang.com/api/v1/knowledge/claims/1138469MCP
get_claim(id=1138469)