SparsePO: Controlling Preference Alignment of LLMs via Sparse Token Masks

• URL: http://arxiv.org/abs/2410.05102v2
• Date: Tue, 8 Oct 2024 15:53:22 GMT
• Title: SparsePO: Controlling Preference Alignment of LLMs via Sparse Token Masks
• Authors: Fenia Christopoulou, Ronald Cardenas, Gerasimos Lampouras, Haitham Bou-Ammar, Jun Wang,
• Abstract summary: We propose a flexible objective termed SparsePO to automatically learn to weight the KL divergence and reward corresponding to each token during Preference Optimization training. We show that our approach assigns meaningful weights to tokens according to the target task, generates more responses with the desired preference and improves reasoning tasks by up to 2 percentage points compared to other token- and response-level PO methods.
• Score: 13.600674179059238
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Preference Optimization (PO) has proven an effective step for aligning language models to human-desired behaviors. Current variants, following the offline Direct Preference Optimization objective, have focused on a strict setting where all tokens are contributing signals of KL divergence and rewards to the loss function. However, human preference is not affected by each word in a sequence equally but is often dependent on specific words or phrases, e.g. existence of toxic terms leads to non-preferred responses. Based on this observation, we argue that not all tokens should be weighted equally during PO and propose a flexible objective termed SparsePO, that aims to automatically learn to weight the KL divergence and reward corresponding to each token during PO training. We propose two different variants of weight-masks that can either be derived from the reference model itself or learned on the fly. Notably, our method induces sparsity in the learned masks, allowing the model to learn how to best weight reward and KL divergence contributions at the token level, learning an optimal level of mask sparsity. Extensive experiments on multiple domains, including sentiment control, dialogue, text summarization and text-to-code generation, illustrate that our approach assigns meaningful weights to tokens according to the target task, generates more responses with the desired preference and improves reasoning tasks by up to 2 percentage points compared to other token- and response-level PO methods.

Related papers

• Not all tokens are created equal: Perplexity Attention Weighted Networks for AI generated text detection [49.15148871877941]
  Next-token distribution outputs offer a theoretically appealing approach for detection of large language models (LLMs) We propose the Perplexity Attention Weighted Network (PAWN), which uses the last hidden states of the LLM and positions to weight the sum of a series of features based on metrics from the next-token distribution across the sequence length. PAWN shows competitive and even better performance in-distribution than the strongest baselines with a fraction of their trainable parameters.
  arXiv  Detail & Related papers  (2025-01-07T17:00:49Z)
• Disentangling Reasoning Tokens and Boilerplate Tokens For Language Model Fine-tuning [46.43130011147807]
  We argue that tokens serving different roles - specifically, reasoning tokens versus boilerplate tokens - differ significantly in importance and learning complexity. We propose a novel Shuffle-Aware Discriminator (SHAD) for adaptive token discrimination. Using SHAD, we propose the Reasoning-highlighted Fine-Tuning (RFT) method, which adaptively emphasizes reasoning tokens during fine-tuning.
  arXiv  Detail & Related papers  (2024-12-19T12:06:24Z)
• Adaptive Dense Reward: Understanding the Gap Between Action and Reward Space in Alignment [33.5805074836187]
  Reinforcement Learning from Human Feedback (RLHF) has proven highly effective in aligning Large Language Models (LLMs) with human preferences. This limitation stems from RLHF's lack of awareness regarding which specific tokens should be reinforced or suppressed. We propose the Adaptive Message-wise RLHF'' method, which robustly applies to various tasks.
  arXiv  Detail & Related papers  (2024-10-23T16:16:15Z)
• TIS-DPO: Token-level Importance Sampling for Direct Preference Optimization With Estimated Weights [73.9088920210495]
  We propose a token-level importance sampling DPO objective named TIS-DPO that assigns importance weights to each token based on its reward. TIS-DPO significantly outperforms various baseline methods on harmlessness and helpfulness alignment and summarization tasks.
  arXiv  Detail & Related papers  (2024-10-06T04:03:00Z)
• TLCR: Token-Level Continuous Reward for Fine-grained Reinforcement Learning from Human Feedback [24.4488286574098]
  We introduce TLCR (Token-Level Continuous Reward) for Reinforcement Learning from Human Feedback. Our proposed TLCR leads to consistent performance improvements over previous sequence-level or token-level discrete rewards.
  arXiv  Detail & Related papers  (2024-07-23T15:27:37Z)
• Orchestrating LLMs with Different Personalizations [28.344891363780576]
  This paper presents a novel approach to aligning large language models (LLMs) with individual human preferences. Given stated preferences along multiple dimensions, such as helpfulness, conciseness, or humor, the goal is to create an LLM without re-training that best adheres to this specification. Starting from specialized expert LLMs, each trained for one particular preference dimension, we propose a black-box method that merges their outputs on a per-token level.
  arXiv  Detail & Related papers  (2024-07-04T22:55:02Z)
• SEP: Self-Enhanced Prompt Tuning for Visual-Language Model [93.94454894142413]
  We introduce a novel approach named Self-Enhanced Prompt Tuning (SEP) SEP explicitly incorporates discriminative prior knowledge to enhance both textual-level and visual-level embeddings. Comprehensive evaluations across various benchmarks and tasks confirm SEP's efficacy in prompt tuning.
  arXiv  Detail & Related papers  (2024-05-24T13:35:56Z)
• Token-level Direct Preference Optimization [8.249403373337024]
  Fine-tuning pre-trained Large Language Models is essential to align them with human values and intentions. We introduce Token-level Direct Preference Optimization (TDPO), a novel approach to align LLMs with human preferences by optimizing policy at the token level.
  arXiv  Detail & Related papers  (2024-04-18T08:49:38Z)
• Relative Preference Optimization: Enhancing LLM Alignment through Contrasting Responses across Identical and Diverse Prompts [95.09994361995389]
  Relative Preference Optimization (RPO) is designed to discern between more and less preferred responses derived from both identical and related prompts. RPO has demonstrated a superior ability to align large language models with user preferences and to improve their adaptability during the training process.
  arXiv  Detail & Related papers  (2024-02-12T22:47:57Z)
• Parameter-Efficient Tuning Helps Language Model Alignment [57.27390187540737]
  Previous works mainly adopt reinforcement learning (RLHF) and direct preference optimization (DPO) with human feedback for alignment. Controllable generation offers more flexibility with regard to data format. Our approach, alignMEnt with parameter-Efficient Tuning (MEET), improves the quality of control tokens.
  arXiv  Detail & Related papers  (2023-10-01T23:27:14Z)
• Effective Unsupervised Domain Adaptation with Adversarially Trained Language Models [54.569004548170824]
  We show that careful masking strategies can bridge the knowledge gap of masked language models. We propose an effective training strategy by adversarially masking out those tokens which are harder to adversarial by the underlying.
  arXiv  Detail & Related papers  (2020-10-05T01:49:47Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.