Unverified50% confidenceBenchmarkExact time
The study uses the Unitree G1 as its platform and achieved a 92.5% success rate in a simulated environment.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
10/26/2026
First Seen
Valid until: 1/24/2027
Sources
Related Entities
Related Claims
UnverifiedOn action appropriateness, the average score across all models reached 95.6%.65% similarUnverifiedGGUF量化后的Q4模型在大多数任务上仍能保留原始模型90%以上的能力65% similarUnverifiedAt the mention level, the model achieved an overall precision of 74.1% and a recall of 70.5%.64% similarUnverifiedIn the benchmark, using the open-source model GLM with TrueForge reduced costs by as much as 75% while maintaining accuracy.64% similarUnverified在GPQA V2版本测试中，Glimmer的幻觉率高达82%64% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1099876API
curl https://kongchang.com/api/v1/knowledge/claims/1099876MCP
get_claim(id=1099876)