Merge branch 'standalone_old_log_prob_support' of https://github.com/meituan-search/verl into standalone_old_log_prob_support

Author: sl-1314

verl/experimental/agent_loop/tool_agent_loop.py

@@ -251,7 +251,7 @@ async def _handle_generating_state(
         if output.log_probs:
             agent_data.response_logprobs += output.log_probs
         if output.old_log_probs:
-            agent_data.response_logprobs += output.old_log_probs
+            agent_data.response_oldlogprobs += output.old_log_probs
         if output.routed_experts is not None:
             agent_data.routed_experts = output.routed_experts