Navigating Simply, Aligning Deeply: Winning Solutions for Mouse vs. AI 2025

TL;DR

This paper presents winning solutions for the NeurIPS 2025 Mouse vs. AI competition, demonstrating that simple architectures can excel in robustness while deeper models improve neural alignment, challenging traditional complexity assumptions.

Contribution

The paper introduces effective, lightweight models for visual robustness and neural alignment, with systematic analysis and insights into model performance and training dynamics.

Findings

Simple architectures achieve high robustness (95.4% score).

Deeper models with more capacity excel in neural alignment.

Optimal training performance occurs around 200K steps.

Abstract

Visual robustness and neural alignment remain critical challenges in developing artificial agents that can match biological vision systems. We present the winning approaches from Team HCMUS_TheFangs for both tracks of the NeurIPS 2025 Mouse vs. AI: Robust Visual Foraging Competition. For Track 1 (Visual Robustness), we demonstrate that architectural simplicity combined with targeted components yields superior generalization, achieving 95.4% final score with a lightweight two-layer CNN enhanced by Gated Linear Units and observation normalization. For Track 2 (Neural Alignment), we develop a deep ResNet-like architecture with 16 convolutional layers and GLU-based gating that achieves top-1 neural prediction performance with 17.8 million parameters. Our systematic analysis of ten model checkpoints trained between 60K to 1.14M steps reveals that training duration exhibits a non-monotonic relationship with performance, with optimal results achieved around 200K steps. Through comprehensive ablation studies and failure case analysis, we provide insights into why simpler architectures excel at visual robustness while deeper models with increased capacity achieve better neural alignment. Our results challenge conventional assumptions about model complexity in visuomotor learning and offer practical guidance for developing robust, biologically-inspired visual agents.