# The Dark Side of Rich Rewards: Understanding and Mitigating Noise in VLM Rewards

arXiv: https://arxiv.org/abs/2409.15922 · HTML: https://mlsanity.com/paper/2409.15922/

## Key facts

- Cited by 7 papers in the map (data through 14 Oct 2026).
- Placed in Embodied and 3D vision-language models, one of 16 directions in Multimodal LLMs / vision-language models.
- Builds on 32 of its 64 references that are in the map.
- First posted on arXiv 24 Sep 2024.

## Placed in

- Multimodal LLMs / vision-language models / [Embodied and 3D vision-language models](https://mlsanity.com/mllm/embodied-and-3d-vision-language-models/)

## Builds on (most cited first, 30 of 32)

- [Learning Transferable Visual Models From Natural Language Supervision](https://mlsanity.com/paper/2103.00020/) (2021)
- [Training language models to follow instructions with human feedback](https://mlsanity.com/paper/2203.02155/) (2022)
- [Proximal Policy Optimization Algorithms](https://mlsanity.com/paper/1707.06347/) (2017)
- [RoFormer: Enhanced Transformer with Rotary Position Embedding](https://mlsanity.com/paper/2104.09864/) (2021)
- [Conservative Q-Learning for Offline Reinforcement Learning](https://mlsanity.com/paper/2006.04779/) (2020)
- [Meta-World: A Benchmark and Evaluation for Multi-Task and Meta Reinforcement Learning](https://mlsanity.com/paper/1910.10897/) (2019)
- [Exploration by Random Network Distillation](https://mlsanity.com/paper/1810.12894/) (2018)
- [CLIPort: What and Where Pathways for Robotic Manipulation](https://mlsanity.com/paper/2109.12098/) (2021)
- [Faith and Fate: Limits of Transformers on Compositionality](https://mlsanity.com/paper/2305.18654/) (2023)
- [Deep Recurrent Q-Learning for Partially Observable MDPs](https://mlsanity.com/paper/1507.06527/) (2015)
- [Is Pessimism Provably Efficient for Offline RL?](https://mlsanity.com/paper/2012.15085/) (2020)
- [Minigrid & Miniworld: Modular & Customizable Reinforcement Learning Environments for Goal-Oriented Tasks](https://mlsanity.com/paper/2306.13831/) (2023)
- [Reinforced Cross-Modal Matching and Self-Supervised Imitation Learning for Vision-Language Navigation](https://mlsanity.com/paper/1811.10092/) (2018)
- [X-CLIP: End-to-End Multi-grained Contrastive Learning for Video-Text Retrieval](https://mlsanity.com/paper/2207.07285/) (2022)
- [Least Ambiguous Set-Valued Classifiers with Bounded Error Levels](https://mlsanity.com/paper/1609.00451/) (2016)
- [BabyAI: A Platform to Study the Sample Efficiency of Grounded Language Learning](https://mlsanity.com/paper/1810.08272/) (2018)
- [Benchmarking the Spectrum of Agent Capabilities](https://mlsanity.com/paper/2109.06780/) (2021)
- [Vision-Language Models are Zero-Shot Reward Models for Reinforcement Learning](https://mlsanity.com/paper/2310.12921/) (2023)
- [Pessimistic Model-based Offline Reinforcement Learning under Partial Coverage](https://mlsanity.com/paper/2107.06226/) (2021)
- [RL-VLM-F: Reinforcement Learning from Vision Language Foundation Model Feedback](https://mlsanity.com/paper/2402.03681/) (2024)
- [Using Natural Language for Reward Shaping in Reinforcement Learning](https://mlsanity.com/paper/1903.02020/) (2019)
- [Language models are not naysayers: An analysis of language models on negation benchmarks](https://mlsanity.com/paper/2306.08189/) (2023)
- [Out of Order: How Important Is The Sequential Order of Words in a Sentence in Natural Language Understanding Tasks?](https://mlsanity.com/paper/2012.15180/) (2020)
- [On the Expressivity of Markov Reward](https://mlsanity.com/paper/2111.00876/) (2021)
- [Vision-Language Models as a Source of Rewards](https://mlsanity.com/paper/2312.09187/) (2023)
- [Heuristic-Guided Reinforcement Learning](https://mlsanity.com/paper/2106.02757/) (2021)
- [An Analysis of Negation in Natural Language Understanding Corpora](https://mlsanity.com/paper/2203.08929/) (2022)
- [PixL2R: Guiding Reinforcement Learning Using Natural Language by Mapping Pixels to Rewards](https://mlsanity.com/paper/2007.15543/) (2020)
- [The Effect of Modeling Human Rationality Level on Learning Rewards from Multiple Feedback Types](https://mlsanity.com/paper/2208.10687/) (2022)
- [FuRL: Visual-Language Models as Fuzzy Rewards for Reinforcement Learning](https://mlsanity.com/paper/2406.00645/) (2024)

## Built on by (most cited first, 7 of 7)

- [Checklists Are Better Than Reward Models For Aligning Language Models](https://mlsanity.com/paper/2507.18624/) (2025)
- [Toward Stable World Models: Measuring and Addressing World Instability in Generative Environments](https://mlsanity.com/paper/2503.08122/) (2025)
- [Discretizing Reward Models](https://mlsanity.com/paper/2606.21795/) (2026)
- [MARVL: Multi-Stage Guidance for Robotic Manipulation via Vision-Language Models](https://mlsanity.com/paper/2602.15872/) (2026)
- [MVR: Multi-view Video Reward Shaping for Reinforcement Learning](https://mlsanity.com/paper/2603.01694/) (2026)
- [Benchmarking LLM Judges for Mobile Agent Evaluation](https://mlsanity.com/paper/2608.11434/) (2026)
- [FOSSIL: Harnessing Feedback on Suboptimal Samples for Data-Efficient Generalisation with Imitation Learning for Embodied Vision-and-Language Tasks](https://mlsanity.com/paper/2510.11307/) (2025)

## Abstract (arXiv)

While Vision-Language Models (VLMs) are increasingly used to generate reward signals for training embodied agents to follow instructions, our research reveals that agents guided by VLM rewards often underperform compared to those employing only intrinsic (exploration-driven) rewards, contradicting expectations set by recent work. We hypothesize that false positive rewards -- instances where unintended trajectories are incorrectly rewarded -- are more detrimental than false negatives. Our analysis confirms this hypothesis, revealing that the widely used cosine similarity metric is prone to false positive reward estimates. To address this, we introduce BiMI ({Bi}nary {M}utual {I}nformation), a novel reward function designed to mitigate noise. BiMI significantly enhances learning efficiency across diverse and challenging embodied navigation environments. Our findings offer a nuanced understanding of how different types of reward noise impact agent learning and highlight the importance of addressing multimodal reward signal noise when training embodied agents