add initial trackio integration for roll.

fix: use Cluster instead of WorkerConfig for dynamic batching dp_size

remove IPA config yaml not needed for OpenReward integration

revert agent_native_env_manager.py to upstream version

added the openreward support.

added OpenRewards

add reference to notable work

Update config name in run_onpolicy_distill_pipeline.sh

fix: disable reward normalization for SWE configs with group_size=1

minor comment.

Add Qwen3.5 ROCK agentic SWE example

Revise RLix description for clarity and detail

fix: MultipleChoiceBoxedRuleRewardWorker returns a zero reward

fix: fix rlvr metrics update

docs: Add Huawei Ascend hardware support doc