07:45
2026-10-15
snipvote.com
artificial-intelligence
GPT-6-astra leads LHTB 2.0 benchmark with 0.69 dense mean reward
GPT-6-astra leads the LHTB 2.0 long-horizon terminal benchmark with a 0.69 dense mean reward, according to an arXiv paper (2610.14146). LHTB 2.0 expands long-horizon terminal evaluation to 77 tasks ac…