# Let's Think in Two Steps: Mitigating Agreement Bias in MLLMs with Self-Grounded Verification

arXiv: https://arxiv.org/abs/2507.11662 · HTML: https://mlsanity.com/paper/2507.11662/

## Key facts

- Cited by 14 papers in the map (data through 14 Oct 2026).
- Placed in Reasoning for agents, planning and code, one of 17 directions in LLM reasoning.
- Builds on 51 of its 71 references that are in the map.
- First posted on arXiv 15 Jul 2025.

## Placed in

- LLM reasoning / [Reasoning for agents, planning and code](https://mlsanity.com/lr/reasoning-for-agents-planning-and-code/)

## Builds on (most cited first, 30 of 51)

- [DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models](https://mlsanity.com/paper/2402.03300/) (2024)
- [Judging LLM-as-a-Judge with MT-Bench and Chatbot Arena](https://mlsanity.com/paper/2306.05685/) (2023)
- [DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning](https://mlsanity.com/paper/2501.12948/) (2025)
- [BLIP-2: Bootstrapping Language-Image Pre-training with Frozen Image Encoders and Large Language Models](https://mlsanity.com/paper/2301.12597/) (2023)
- [Self-Consistency Improves Chain of Thought Reasoning in Language Models](https://mlsanity.com/paper/2203.11171/) (2022)
- [Let's Verify Step by Step](https://mlsanity.com/paper/2305.20050/) (2023)
- [Voyager: An Open-Ended Embodied Agent with Large Language Models](https://mlsanity.com/paper/2305.16291/) (2023)
- [WebArena: A Realistic Web Environment for Building Autonomous Agents](https://mlsanity.com/paper/2307.13854/) (2023)
- [Scaling LLM Test-Time Compute Optimally can be More Effective than Scaling Model Parameters](https://mlsanity.com/paper/2408.03314/) (2024)
- [AgentBench: Evaluating LLMs as Agents](https://mlsanity.com/paper/2308.03688/) (2023)
- [Kimi k1.5: Scaling Reinforcement Learning with LLMs](https://mlsanity.com/paper/2501.12599/) (2025)
- [Towards Understanding Sycophancy in Language Models](https://mlsanity.com/paper/2310.13548/) (2023)
- [Large Language Models Cannot Self-Correct Reasoning Yet](https://mlsanity.com/paper/2310.01798/) (2023)
- [Reasoning with Language Model is Planning with World Model](https://mlsanity.com/paper/2305.14992/) (2023)
- [Multimodal Chain-of-Thought Reasoning in Language Models](https://mlsanity.com/paper/2302.00923/) (2023)
- [What Matters in Learning from Offline Human Demonstrations for Robot Manipulation](https://mlsanity.com/paper/2108.03298/) (2021)
- [UI-TARS: Pioneering Automated GUI Interaction with Native Agents](https://mlsanity.com/paper/2501.12326/) (2025)
- [Language Agent Tree Search Unifies Reasoning Acting and Planning in Language Models](https://mlsanity.com/paper/2310.04406/) (2023)
- [Monitoring Reasoning Models for Misbehavior and the Risks of Promoting Obfuscation](https://mlsanity.com/paper/2503.11926/) (2025)
- [Agent Workflow Memory](https://mlsanity.com/paper/2409.07429/) (2024)
- [MLLM-as-a-Judge: Assessing Multimodal LLM-as-a-Judge with Vision-Language Benchmark](https://mlsanity.com/paper/2402.04788/) (2024)
- [Justice or Prejudice? Quantifying Biases in LLM-as-a-Judge](https://mlsanity.com/paper/2410.02736/) (2024)
- [Robotic Control via Embodied Chain-of-Thought Reasoning](https://mlsanity.com/paper/2407.08693/) (2024)
- [Robots That Ask For Help: Uncertainty Alignment for Large Language Model Planners](https://mlsanity.com/paper/2307.01928/) (2023)
- [Set-of-Mark Prompting Unleashes Extraordinary Visual Grounding in GPT-4V](https://mlsanity.com/paper/2310.11441/) (2023)
- [Physics of Language Models: Part 3.1, Knowledge Storage and Extraction](https://mlsanity.com/paper/2309.14316/) (2023)
- [DigiRL: Training In-The-Wild Device-Control Agents with Autonomous Reinforcement Learning](https://mlsanity.com/paper/2406.11896/) (2024)
- [Spurious Rewards: Rethinking Training Signals in RLVR](https://mlsanity.com/paper/2506.10947/) (2025)
- [On the Effects of Data Scale on UI Control Agents](https://mlsanity.com/paper/2406.03679/) (2024)
- [LLMs Can't Plan, But Can Help Planning in LLM-Modulo Frameworks](https://mlsanity.com/paper/2402.01817/) (2024)

## Built on by (most cited first, 14 of 14)

- [Mobile-Agent-v3.5: Multi-platform Fundamental GUI Agents](https://mlsanity.com/paper/2602.16855/) (2026)
- [WALT: Web Agents that Learn Tools](https://mlsanity.com/paper/2510.01524/) (2025)
- [When to Act, Ask, or Learn: Uncertainty-Aware Policy Steering](https://mlsanity.com/paper/2602.22474/) (2026)
- [VLAA-GUI: Knowing When to Stop, Recover, and Search, A Modular Framework for GUI Automation](https://mlsanity.com/paper/2604.21375/) (2026)
- [Structured Distillation of Web Agent Capabilities Enables Generalization](https://mlsanity.com/paper/2604.07776/) (2026)
- [StateAct: Program State, before Pixels, for Long-Horizon Computer-Use Agents](https://mlsanity.com/paper/2607.22798/) (2026)
- [Verification Mirage: Mapping the Reliability Boundary of Self-Verification in Medical VQA](https://mlsanity.com/paper/2605.10850/) (2026)
- [PANDO: Efficient Multimodal AI Agents via Online Skill Distillation](https://mlsanity.com/paper/2605.24785/) (2026)
- [ERRAND: Budgeted Maintenance of Agent Memory](https://mlsanity.com/paper/2609.29545/) (2026)
- [Diagnosing and Mitigating Perception-Decision Misalignment in Omni-LLMs via Modality Subspace Activation](https://mlsanity.com/paper/2608.14655/) (2026)
- [Don't Read the Log: Execution Traces Contaminate Verifiers in Video-Generation Agents](https://mlsanity.com/paper/2609.28564/) (2026)
- [Probe to Act: Elevating Browser-Use Agent via Active Visual Probing](https://mlsanity.com/paper/2609.33646/) (2026)
- [RSure-Agent: Reliable Use of Tool Observations for Remote Sensing Agents](https://mlsanity.com/paper/2610.04836/) (2026)
- [Shared-Roadmap Generation and Evaluator for Multi-Agent Path Planning Using Heterogeneous Graph Neural Network](https://mlsanity.com/paper/2610.09034/) (2026)

## Abstract (arXiv)

Verifiers--functions assigning rewards to agent behavior--have been key to AI progress in math, code, and games. However, extending gains to domains without clear-cut success criteria remains a challenge: while humans can recognize desired outcomes, translating this intuition into scalable rules is nontrivial. Multimodal LLMs (MLLMs) offer a promising solution, given their world knowledge, human-preference alignment, and reasoning capabilities. We evaluate MLLM verifiers across web navigation, computer use, and robotics, spanning 13+ models, 28+ designs, and thousands of trajectories from diverse agents. We identify a critical limitation: a strong tendency for MLLMs to over-validate agent behavior--a phenomenon we term agreement bias. This bias is pervasive, resilient to test-time scaling, and can harm applications relying on MLLM judgments/rewards (e.g., self-improvement, steering, online supervision). We discuss several considerations for evaluating and designing MLLM verifiers, and introduce SGV, a lightweight method that better leverages their capabilities by modulating (un)conditional generation. First, an MLLM is elicited to generate broad priors about desired behavior, independent of the data under evaluation. Then, conditioned on self-generated priors, it reasons over and evaluates a candidate trajectory. Our methods yield more human-aligned verifiers, improving failure detection by 25pp and accuracy by 14pp. In self-improvement and online supervision, they boost task completion of a GUI specialist in OSWorld, a diffusion policy in robomimic, and a ReAct agent in VisualWebArena--surpassing the previous state of the art by 20pp. As a byproduct, we release an update of VisualWebArena featuring strong agent baselines, more human-aligned oracles, container parallelism with high fidelity and proper resets, >10x speedups, and VWA-Lite, a 1/3 subset with comparable evaluation fidelity.