@article{thedarkroomintherewardchanneldensepredic,
  title = {The Dark Room in the Reward Channel: Dense Prediction Rewards Collapse GRPO-Trained LLM Agents -- and The Channel, Not the Content, Decides What Works},
  author = {Yu Wang},
  year = {2026},
  eprint = {2607.21273},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2607.21273},
}