Fixing GRPO training collapse in long-horizon multi-tool agents. A lightweight PRM-Lite + LATA joint approach achieves +37% over vanilla GRPO on τ-bench airline (50-task, multi-turn).
Fixing GRPO training collapse in long-horizon multi-tool agents. A lightweight PRM-Lite + LATA joint approach achieves +37% over vanilla GRPO on τ-bench airline (50-task, multi-turn).
Marketplace
Independent
Category
automation
More like this
Browse automation agents →
Glass Ripple
The hero effect for the AI agent era — GPU-accelerated glass ripple with CRT retro aesthetics with a lot brand icons, 11-pass WebGL2 shader pipeline
Free
automationZapier AI
Automate workflows with natural language
$19/mo
automationGoogle Workspace AI
Gemini AI across Gmail, Docs, Sheets, and Meet
$30/mo
automationn8n
Open-source workflow automation with AI nodes
$20/mo