# Shortcuts in the Tail: Debiasing via Post-Hoc Spectral Compression of Fine-Tuning Updates

arXiv: https://arxiv.org/abs/2606.07596 · HTML: https://mlsanity.com/paper/2606.07596/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 18 of its 24 references that are in the map.
- First posted on arXiv 29 May 2026.

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

- [A Broad-Coverage Challenge Corpus for Sentence Understanding through Inference](https://mlsanity.com/paper/1704.05426/) (2017)
- [FEVER: a large-scale dataset for Fact Extraction and VERification](https://mlsanity.com/paper/1803.05355/) (2018)
- [Distributionally Robust Neural Networks for Group Shifts: On the Importance of Regularization for Worst-Case Generalization](https://mlsanity.com/paper/1911.08731/) (2019)
- [Editing Models with Task Arithmetic](https://mlsanity.com/paper/2212.04089/) (2022)
- [Right for the Wrong Reasons: Diagnosing Syntactic Heuristics in Natural Language Inference](https://mlsanity.com/paper/1902.01007/) (2019)
- [Nuanced Metrics for Measuring Unintended Bias with Real Data for Text Classification](https://mlsanity.com/paper/1903.04561/) (2019)
- [PAWS: Paraphrase Adversaries from Word Scrambling](https://mlsanity.com/paper/1904.01130/) (2019)
- [SVD-LLM: Truncation-aware Singular Value Decomposition for Large Language Model Compression](https://mlsanity.com/paper/2403.07378/) (2024)
- [Language model compression with weighted low-rank factorization](https://mlsanity.com/paper/2207.00112/) (2022)
- [Mechanistically analyzing the effects of fine-tuning on procedurally defined tasks](https://mlsanity.com/paper/2311.12786/) (2023)
- [SVD-LLM V2: Optimizing Singular Value Truncation for Large Language Model Compression](https://mlsanity.com/paper/2503.12340/) (2025)
- [Explore Spurious Correlations at the Concept Level in Language Models for Text Classification](https://mlsanity.com/paper/2311.08648/) (2023)
- [RaVL: Discovering and Mitigating Spurious Correlations in Fine-Tuned Vision-Language Models](https://mlsanity.com/paper/2411.04097/) (2024)
- [Safety Mirage: How Spurious Correlations Undermine VLM Safety Fine-Tuning and Can Be Mitigated by Machine Unlearning](https://mlsanity.com/paper/2503.11832/) (2025)
- [Escaping the SpuriVerse: Can Large Vision-Language Models Generalize Beyond Seen Spurious Correlations?](https://mlsanity.com/paper/2506.18322/) (2025)
- [Assessing Robustness to Spurious Correlations in Post-Training Language Models](https://mlsanity.com/paper/2505.05704/) (2025)
- [When Bias Pretends to Be Truth: How Spurious Correlations Undermine Hallucination Detection in LLMs](https://mlsanity.com/paper/2511.07318/) (2025)
- [LoRA Users Beware: A Few Spurious Tokens Can Manipulate Your Finetuned Model](https://mlsanity.com/paper/2506.11402/) (2025)

## Abstract (arXiv)

Fine-tuning often introduces spurious correlations alongside task knowledge, causing systematic failures on underrepresented groups. Existing mitigations require retraining, group labels, or curated counterfactual data. We show a simple post-hoc intervention reduces shortcut reliance without any of these: truncating the tail of the SVD of Delta W = Wft - Wbase reduces the spurious-group gap while preserving task accuracy. Across three instruction-tuned models (0.5B--7B) and four classification benchmarks, top-k truncation reduces the gap on every cell at <2 pp accuracy loss, by up to 5× on CivilComments. We propose this works because the shortcut response sits in the tail of the singular ordering of Delta W, a claim about how truncation behaves rather than about the raw singular values, which are broadly distributed and look the same across all four datasets. A controlled boundary case in which fine-tuning has only a shortcut to learn shows the predicted FT-to-base collapse, and bottom-/random-k and matched-rank LoRA controls rule out generic low-rank approximation and rank-constrained training as the explanation. We read this as preliminary evidence that the singular basis of Delta W is a useful coordinate system for studying what fine-tuning has learned.