# EGA: Adapting Frozen Encoders for Vector Search with Bounded Out-of-Distribution Degradation

arXiv: https://arxiv.org/abs/2605.05674 · HTML: https://mlsanity.com/paper/2605.05674/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 30 of its 43 references that are in the map.
- First posted on arXiv 7 May 2026.

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

- [Learning Transferable Visual Models From Natural Language Supervision](https://mlsanity.com/paper/2103.00020/) (2021)
- [DINOv2: Learning Robust Visual Features without Supervision](https://mlsanity.com/paper/2304.07193/) (2023)
- [Representation Learning with Contrastive Predictive Coding](https://mlsanity.com/paper/1807.03748/) (2018)
- [Adding Conditional Control to Text-to-Image Diffusion Models](https://mlsanity.com/paper/2302.05543/) (2023)
- [Squeeze-and-Excitation Networks](https://mlsanity.com/paper/1709.01507/) (2017)
- [Parameter-Efficient Transfer Learning for NLP](https://mlsanity.com/paper/1902.00751/) (2019)
- [Sigmoid Loss for Language Image Pre-Training](https://mlsanity.com/paper/2303.15343/) (2023)
- [Billion-scale similarity search with GPUs](https://mlsanity.com/paper/1702.08734/) (2017)
- [Learning to Prompt for Vision-Language Models](https://mlsanity.com/paper/2109.01134/) (2021)
- [SimCSE: Simple Contrastive Learning of Sentence Embeddings](https://mlsanity.com/paper/2104.08821/) (2021)
- [Visual Prompt Tuning](https://mlsanity.com/paper/2203.12119/) (2022)
- [Conditional Prompt Learning for Vision-Language Models](https://mlsanity.com/paper/2203.05557/) (2022)
- [Understanding Contrastive Representation Learning through Alignment and Uniformity on the Hypersphere](https://mlsanity.com/paper/2005.10242/) (2020)
- [Efficient and robust approximate nearest neighbor search using Hierarchical Navigable Small World graphs](https://mlsanity.com/paper/1603.09320/) (2016)
- [Tip-Adapter: Training-free Adaption of CLIP for Few-shot Classification](https://mlsanity.com/paper/2207.09519/) (2022)
- [Compositional Entailment Learning for Hyperbolic Vision-Language Models](https://mlsanity.com/paper/2410.06912/) (2024)
- [SPANN: Highly-efficient Billion-scale Approximate Nearest Neighbor Search](https://mlsanity.com/paper/2111.08566/) (2021)
- [Rethinking Graph Masked Autoencoders through Alignment and Uniformity](https://mlsanity.com/paper/2402.07225/) (2024)
- [Worst-case Performance of Popular Approximate Nearest Neighbor Search Implementations: Guarantees and Limitations](https://mlsanity.com/paper/2310.19126/) (2023)
- [I-Con: A Unifying Framework for Representation Learning](https://mlsanity.com/paper/2504.16929/) (2025)
- [boldsymbolλ-Orthogonality Regularization for Compatible Representation Learning](https://mlsanity.com/paper/2509.16664/) (2025)
- [Navigable Graphs for High-Dimensional Nearest Neighbor Search: Constructions and Limits](https://mlsanity.com/paper/2405.18680/) (2024)
- [Improve Representation for Imbalanced Regression through Geometric Constraints](https://mlsanity.com/paper/2503.00876/) (2025)
- [The Geometry of Representational Failures in Vision Language Models](https://mlsanity.com/paper/2602.07025/) (2026)
- [LoRANN: Low-Rank Matrix Factorization for Approximate Nearest Neighbor Search](https://mlsanity.com/paper/2410.18926/) (2024)
- [A Principled Framework for Multi-View Contrastive Learning](https://mlsanity.com/paper/2507.06979/) (2025)
- [PCA++: How Uniformity Induces Robustness to Background Noise in Contrastive Learning](https://mlsanity.com/paper/2511.12278/) (2025)
- [PROTOCOL: Partial Optimal Transport-enhanced Contrastive Learning for Imbalanced Multi-view Clustering](https://mlsanity.com/paper/2506.12408/) (2025)
- [UniCon: Unified Framework for Efficient Contrastive Alignment via Kernels](https://mlsanity.com/paper/2604.16678/) (2026)
- [An Information-Theoretic Regularizer for Lossy Neural Image Compression](https://mlsanity.com/paper/2411.16727/) (2024)

## Abstract (arXiv)

Vector search systems built on frozen vision encoders face queries from unseen classes at deployment, yet existing adapter training collapses under this shift: high-capacity adapters with global contrastive losses silently reassign unseen-class samples to wrong seen-class clusters, dropping worst-case Label Precision by over 40 points below the frozen baseline in our tests. We propose Euclidean Geodesic Alignment (EGA), a residual adapter that couples three principles: zero initialization, local triplet loss, and hypersphere projection. These collectively induce a self-limiting dynamic: triplets that already satisfy a small margin stop producing gradients, so the adapter automatically stops updating where the local geometry is already correct. Our experiments show that at convergence 96.5% of triplets are gradient-free, leaving unseen-class regions largely untouched while still enabling full-capacity refinement of seen classes. Across five diverse out-of-distribution (OOD) benchmarks, EGA achieves the highest worst-case Label Precision on the four primary splits and a consistent improvement on the fifth. The design also transfers to stronger backbones in addition to CLIP, and we provide an analytical justification linking gradient sparsity to bounded OOD perturbation.