Search papers, labs, and topics across Lattice.
This paper introduces Weightless Fine-Tuning (WFT), a novel method for personalizing large language models (LLMs) that avoids the high costs of supervised fine-tuning by operating in logit space during decoding. WFT computes supervised residuals from an author's training data and uses a cross-prefix transport operator to effectively propagate these adjustments to the current prompt, achieving significant performance improvements. On three LaMP personalization benchmarks, WFT not only matches or exceeds the performance of traditional supervised fine-tuning but does so with less than 7% of the computational cost, demonstrating its efficiency and effectiveness.
Achieving SFT-level performance with less than 7% of the computation, Weightless Fine-Tuning revolutionizes how we personalize LLMs without costly weight updates.
Supervised fine-tuning (SFT) is a standard approach for adapting LLMs to a target distribution, but in settings such as personalization, where each author requires separate weight access, optimization, storage, and retraining, its costs become prohibitive. We propose Weightless Fine-Tuning (WFT), a training-free decoding-time method that approximates the distributional effect of SFT without weight updates. WFT computes supervised residuals on an author's training sequence and transports them to the current prompt through a cross-prefix transport operator estimated from dropout-induced cross-covariance. The operator captures how a perturbation at one context propagates to predictions at another, replacing gradient-based parameter updates with logit-space corrections. On three LaMP personalization benchmarks, WFT achieves the best average performance across datasets, matches or exceeds SFT on individual tasks, and outperforms other lightweight baselines on average. In a budget-controlled comparison, WFT approaches SFT performance using less than 7% of the effective computation. Logit-level analysis shows a cosine similarity of 0.875 between the logit shifts induced by WFT and SFT over 95% of the next-token probability mass, suggesting that WFT captures the distributional effect of supervised adaptation without modifying model weights.
