# Useful Features, Backward Scores: OOD in Language-Model Trajectories

arXiv: https://arxiv.org/abs/2605.00269 · HTML: https://mlsanity.com/paper/2605.00269/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 18 of its 25 references that are in the map.
- First posted on arXiv 30 Apr 2026.

## Builds on (most cited first, 18 of 18)

- [A Simple Unified Framework for Detecting Out-of-Distribution Samples and Adversarial Attacks](https://mlsanity.com/paper/1807.03888/) (2018)
- [Out-of-Distribution Detection with Deep Nearest Neighbors](https://mlsanity.com/paper/2204.06507/) (2022)
- [Likelihood Ratios for Out-of-Distribution Detection](https://mlsanity.com/paper/1906.02845/) (2019)
- [Do Deep Generative Models Know What They Don't Know?](https://mlsanity.com/paper/1810.09136/) (2018)
- [TweetEval: Unified Benchmark and Comparative Evaluation for Tweet Classification](https://mlsanity.com/paper/2010.12421/) (2020)
- [Nuanced Metrics for Measuring Unintended Bias with Real Data for Text Classification](https://mlsanity.com/paper/1903.04561/) (2019)
- [ToxicChat: Unveiling Hidden Challenges of Toxicity Detection in Real-World User-AI Conversation](https://mlsanity.com/paper/2310.17389/) (2023)
- [Input complexity and out-of-distribution detection with likelihood-based generative models](https://mlsanity.com/paper/1909.11480/) (2019)
- [Types of Out-of-Distribution Texts and How to Detect Them](https://mlsanity.com/paper/2109.06827/) (2021)
- [Density of States Estimation for Out-of-Distribution Detection](https://mlsanity.com/paper/2006.09273/) (2020)
- [Mahalanobis++: Improving OOD Detection via Feature Normalization](https://mlsanity.com/paper/2505.18032/) (2025)
- [Revisiting Uncertainty Quantification Evaluation in Language Models: Spurious Interactions with Response Length Bias Results](https://mlsanity.com/paper/2504.13677/) (2025)
- [Embedding Trajectory for Out-of-Distribution Detection in Mathematical Reasoning](https://mlsanity.com/paper/2405.14039/) (2024)
- [NLP-ADBench: NLP Anomaly Detection Benchmark](https://mlsanity.com/paper/2412.04784/) (2024)
- [Efficient Hallucination Detection for LLMs Using Uncertainty-Aware Attention Heads](https://mlsanity.com/paper/2505.20045/) (2025)
- [D2HScore: Reasoning-Aware Hallucination Detection via Semantic Breadth and Depth Analysis in LLMs](https://mlsanity.com/paper/2509.11569/) (2025)
- [A Geometry-Based View of Mahalanobis OOD Detection](https://mlsanity.com/paper/2510.15202/) (2025)
- [TrajGuard: Streaming Hidden-state Trajectory Detection for Decoding-time Jailbreak Defense](https://mlsanity.com/paper/2604.07727/) (2026)

## Abstract (arXiv)

Out-of-distribution (OOD) detectors prioritize inputs for closer inspection. Yet features that distinguish input groups need not yield a useful anomaly ranking. We analyze this gap in language-model trajectories under text-length control and fixed score directions. On Spam development data, an input adaptation of D^2HScore falls from raw AUROC 0.919 to 0.530 after length matching. On length-matched, held-out HateSpeech inputs, the same features yield AUROC 0.644 for a labeled linear classifier but 0.444 for an ID-fitted distance score. ToxicChat shows the same contrast. Feature-selection and backbone controls retain the main reversal pattern. Frozen Civil Comments and TweetEval irony tests also reverse (0.467 and 0.435), extending the finding beyond toxicity. In these contrasts, anomalous groups have farther centers but tighter spread. A labeled, fixed-center feature-space intervention changes rankings: equalizing spread helps some tasks and harms others. OOD evaluation must check the chosen score's ranking even when its features distinguish the classes.