A noise-corrected VLM evaluator reward for PPO improves GUI agent success rates by 12.6 percentage points over zero-shot and 5.1 points over raw evaluator rewards across desktop benchmarks.
Reinforcement learning with perturbed rewards.arXiv preprint arXiv:1810.01032,
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.AI 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Reinforcement Learning for Computer-Use Agents with Autonomous Evaluation
A noise-corrected VLM evaluator reward for PPO improves GUI agent success rates by 12.6 percentage points over zero-shot and 5.1 points over raw evaluator rewards across desktop benchmarks.