Search results

The actor/entropy loss appears somewhat abnormal—it's too large. We launched DeepSWE training based on the Qwen3-4B model, using the following script:

Status: Open.

#391 In rllm-org/rllm;

· Cesilina opened on Feb 9, 2026

There may be some compatibility conflicts between Transformers 5.0 and older vLLM.

Status: Open.

#388 In rllm-org/rllm;

· SyncLionPaw opened on Feb 3, 2026

[Bug] Biased pass@1 metrics due to silent episode dropping

Status: Open.

#382 In rllm-org/rllm;

· XHMY opened on Jan 30, 2026

Supporting deep agents and/or automatically generate sub-agents without predefined flow

Status: Open.

#380 In rllm-org/rllm;

· Wonder1905 opened on Jan 27, 2026

verl update 0.7.0

Status: Open.

#379 In rllm-org/rllm;

· joannanananana opened on Jan 26, 2026

Kind Kubernetes settings

Status: Open.

#378 In rllm-org/rllm;

· bluesky983 opened on Jan 23, 2026

'str' object has no attribute 'get'

Status: Open.

#364 In rllm-org/rllm;

· xuyeliu opened on Jan 11, 2026

How to implement trajectory-level masking for "invalid turns" (SimpleTIR) in rllm?

Status: Open.

#355 In rllm-org/rllm;

· BabelTower opened on Dec 23, 2025

actor_rollout_ref.actor.loss_agg_mode=seq-mean-token-sum in examples/swe/train_deepswe_32b.sh

Status: Open.

#354 In rllm-org/rllm;

· p81sunshine opened on Dec 21, 2025

[Warning] When assemble steps, detect the trajectory not accumulative

bugSomething isn't workingSomething isn't working

Status: Open.

#352 In rllm-org/rllm;

· kxfan2002 opened on Dec 17, 2025

ppo_mini_batch_size for Non-Cumulative Agents

questionFurther information is requestedFurther information is requested

Status: Open.

#350 In rllm-org/rllm;

· huseyinatahaninan opened on Dec 16, 2025

Strands VerlEngine support

Status: Open.

#333 In rllm-org/rllm;

· austinmw opened on Dec 2, 2025