Add qwen3_vl_moe to VLM registry (#2196)

chore: bump verifiers to latest main (d3c830c) (#2186)

add `all` optional-dependency group that combines all extras (#2184)

fix: always log per-task error and empty rollout metrics (even when 0) (#2185)

feat: improve monitor-run skill (#2179)

clean up some metrics (#2163)

Context parallel for DSA (#2173)

Revert "vllm 0.18.0 (#2141)" (#2177)

fix: increase env server shutdown timeout for sandbox cleanup (#2175)

Another iteration on the DP hang (#2168)

fix: update CI Docker image to CUDA 13.0 for deep-gemm compatibility (#2174)

Point vllm-router at PrimeIntellect-ai/router v0.1.13 (#2167)

chore: simplify tmux script windows (#2170)

fix: kill full process trees on rl cleanup (#2158)

feat: set process titles on all entrypoints (#2159)