Gespeichert in:
| Hauptverfasser: | , , , |
|---|---|
| Format: | Preprint |
| Veröffentlicht: |
2025
|
| Schlagworte: | |
| Online-Zugang: | https://arxiv.org/abs/2510.23557 |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| _version_ | 1866914522598473728 |
|---|---|
| author | Réveillard, William Saketos, Vasileios Proutiere, Alexandre Combes, Richard |
| author_facet | Réveillard, William Saketos, Vasileios Proutiere, Alexandre Combes, Richard |
| contents | Training or fine-tuning large language model (LLM)-based systems often requires costly human feedback, yet there is limited understanding of how to minimize such intervention while maintaining strong error guarantees. We study this problem for LLM-based classification systems in an active learning framework: an agent sequentially labels $d$-dimensional query embeddings drawn i.i.d. from an unknown distribution by either calling a costly expert or guessing with no feedback, with the goal of minimizing regret relative to an oracle with free expert access. When the horizon $T$ is at least exponential in the embedding dimension $d$, the geometry of the class regions can be learned. In this regime, we propose the Conservative Hull-based Classifier (CHC), which maintains convex hulls of expert-labeled queries and calls the expert when a query lands outside all known hulls. CHC attains $\mathcal{O}(\log^d T)$ regret in $T$ and is minimax optimal for $d=1$. Otherwise, the geometry cannot be reliably learned in general. We show that for queries drawn from a subgaussian mixture and $T \le e^d$, a Center-based Classifier (CC) achieves regret proportional to $N\log{N}$ where $N$ is the number of labels. To bridge these regimes, we introduce the Generalized Hull-based Classifier (GHC), a practical extension of CHC that enables more aggressive guessing via a tunable parameter. Our approach is validated on real-world question-answering datasets using state-of-the-art text embedding models. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2510_23557 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Minimizing Human Intervention in Online Classification Réveillard, William Saketos, Vasileios Proutiere, Alexandre Combes, Richard Machine Learning Training or fine-tuning large language model (LLM)-based systems often requires costly human feedback, yet there is limited understanding of how to minimize such intervention while maintaining strong error guarantees. We study this problem for LLM-based classification systems in an active learning framework: an agent sequentially labels $d$-dimensional query embeddings drawn i.i.d. from an unknown distribution by either calling a costly expert or guessing with no feedback, with the goal of minimizing regret relative to an oracle with free expert access. When the horizon $T$ is at least exponential in the embedding dimension $d$, the geometry of the class regions can be learned. In this regime, we propose the Conservative Hull-based Classifier (CHC), which maintains convex hulls of expert-labeled queries and calls the expert when a query lands outside all known hulls. CHC attains $\mathcal{O}(\log^d T)$ regret in $T$ and is minimax optimal for $d=1$. Otherwise, the geometry cannot be reliably learned in general. We show that for queries drawn from a subgaussian mixture and $T \le e^d$, a Center-based Classifier (CC) achieves regret proportional to $N\log{N}$ where $N$ is the number of labels. To bridge these regimes, we introduce the Generalized Hull-based Classifier (GHC), a practical extension of CHC that enables more aggressive guessing via a tunable parameter. Our approach is validated on real-world question-answering datasets using state-of-the-art text embedding models. |
| title | Minimizing Human Intervention in Online Classification |
| topic | Machine Learning |
| url | https://arxiv.org/abs/2510.23557 |