Safety
Toward RL on mixed-precision, mixed-accelerator samplers
We trained a model on agentic coding tasks using fully asynchronous RL and mixed-precision, mixed-accelerator samplers. With off-policy correction, training stayed stable through convergence, on par w
We trained a model on agentic coding tasks using fully asynchronous RL and mixed-precision, mixed-accelerator samplers. With off-policy correction, training stayed stable through convergence, on par with homogeneous samplers, at a lower cost.
Related
- Missing Old Logits in Asynchronous Agentic RL: Semantic Mismatch and Repair Methods for Off-Policy Correction
- STAPO: Selective Trajectory-Aware Policy Optimization for LLM Agent Training
- Stabilizing Off-Policy Training for Long-Horizon LLM Agent via Turn-Level Importance Sampling and Clipping-Triggered Normalization
Source: Modal Blog | 2026-10-13