What question did this study set out to answer?

This research aims to enhance out-of-distribution detection by shifting focus from output signals to internal dynamics in neural networks.

January 23, 2026Open Access

DeepDrift / ODD: Hidden-State Trajectory Monitoring for Robustness in Vision, Language, and Control

Key Points

This research aims to enhance out-of-distribution detection by shifting focus from output signals to internal dynamics in neural networks.
Developed Semantic Velocity as a measure of instability in hidden states.
Introduced Optical Depth Dynamics (ODD) for analyzing hidden-state drift across various dimensions.
Evaluated the framework on multiple AI models across vision, language, and control tasks.
Identified distinct drift patterns like Global Collapse in vision architectures and Semantic Tremor in language models.
Detected that high-velocity drift in models can signal impending output failure.
Validated approach on renowned AI architectures, demonstrating low computational overhead.

Abstract

Modern neural networks achieve strong performance on curated benchmarks yet fail unpredictably under distributional shift. Classical out-of-distribution detection relies on output-level confidence signals (e.g., maximum softmax probability; Hendrycks & Gimpel, ICLR 2017). However, for RLHF-tuned large language models and PPO-trained reinforcement learning agents, output confidence is often a lagging indicator: models may remain confident while internally unstable or hallucinating. This work proposes a shift from static output monitoring to internal dynamical diagnostics. We introduce Semantic Velocity, defined as the norm of consecutive hidden-state differences, as a leading indicator of representational instability that often precedes output-level failure. We present Optical Depth Dynamics (ODD), a diagnostic framework that analyzes hidden-state drift across depth—spatial depth for vision models and temporal depth for language and control. ODD provides interpretable failure signatures across modalities: • Vision: Spatial depth analysis reveals characteristic drift patterns such as Global Collapse in ViT-style architectures and Avalanche Effects in CNNs.• Language: Temporal analysis of hidden-state trajectories detects Semantic Tremor: factual generation exhibits stable, low-velocity dynamics, while hallucination is associated with high-variance drift detectable several tokens before completion.• Control: Phase-portrait analysis of reinforcement learning policies identifies Policy Panic regimes in which elevated velocity reliably precedes reward collapse (p < 0.001). All experiments follow explicit statistical protocols, including train/validation/test splits for threshold selection, bootstrap confidence intervals, and effect-size reporting. The approach is validated on Qwen-2.5-7B, TinyLlama-1.1B, LunarLander-v3, and standard vision benchmarks. The open-source library DeepDrift (pip install deepdrift) implements these diagnostics with approximately 1% computational overhead and is intended as a practical monitoring tool for robustness analysis rather than a formal physical theory.

Read Full Paperexternally

Ask AI

Helpful

Bookmark

View Full Paper