Fast Best-of-N Decoding via Speculative Rejection

• URL: http://arxiv.org/abs/2410.20290v2
• Date: Thu, 31 Oct 2024 18:27:13 GMT
• Title: Fast Best-of-N Decoding via Speculative Rejection
• Abstract summary: Inference-time alignment methods avoid the complex post-training step. Best-of-N requires vastly more resources at inference time than standard decoding strategies. We introduce Speculative Rejection, a computationally-viable inference-time alignment algorithm.
• Score: 49.11955026456773
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: The safe and effective deployment of Large Language Models (LLMs) involves a critical step called alignment, which ensures that the model's responses are in accordance with human preferences. Prevalent alignment techniques, such as DPO, PPO and their variants, align LLMs by changing the pre-trained model weights during a phase called post-training. While predominant, these post-training methods add substantial complexity before LLMs can be deployed. Inference-time alignment methods avoid the complex post-training step and instead bias the generation towards responses that are aligned with human preferences. The best-known inference-time alignment method, called Best-of-N, is as effective as the state-of-the-art post-training procedures. Unfortunately, Best-of-N requires vastly more resources at inference time than standard decoding strategies, which makes it computationally not viable. In this work, we introduce Speculative Rejection, a computationally-viable inference-time alignment algorithm. It generates high-scoring responses according to a given reward model, like Best-of-N does, while being between 16 to 32 times more computationally efficient.

Related papers

• Direct Preference Optimization with Rating Information: Practical Algorithms and Provable Gains [67.71020482405343]
  We study how to design algorithms that can leverage additional information in the form of rating gap.<n>We present new algorithms that can achieve faster statistical rates than DPO in presence of accurate rating gap information.
  arXiv  Detail & Related papers  (2026-01-31T08:38:21Z)
• STARS: Segment-level Token Alignment with Rejection Sampling in Large Language Models [14.772543098527786]
  We propose STARS: Segment-level Token Alignment with Rejection Sampling.<n>It steers model generation by iteratively sampling, scoring, and rejecting/accepting short, fixed-size token segments.<n>Our work establishes granular, reward-guided sampling as a generalizable, robust, and efficient alternative to traditional fine-tuning and full-sequence ranking methods.
  arXiv  Detail & Related papers  (2025-11-05T19:47:58Z)
• Kad: A Framework for Proxy-based Test-time Alignment with Knapsack Approximation Deferral [6.949966663998242]
  Large language models (LLM) still require further alignment to adhere to downstream task requirements and stylistic preferences.<n>As LLMs continue to scale in terms of size, the computational cost of alignment procedures increase prohibitively.<n>We propose a novel approach to circumvent these costs via proxy-based test-time alignment.
  arXiv  Detail & Related papers  (2025-10-30T21:38:45Z)
• PITA: Preference-Guided Inference-Time Alignment for LLM Post-Training [9.093854840532062]
  PITA is a novel framework that integrates preference feedback directly into the LLM's token generation.<n> PITA learns a small preference-based guidance policy to modify token probabilities at inference time without fine-tuning.<n>We evaluate PITA across diverse tasks, including mathematical reasoning and sentiment classification.
  arXiv  Detail & Related papers  (2025-07-26T21:46:32Z)
• Can Prompt Difficulty be Online Predicted for Accelerating RL Finetuning of Reasoning Models? [62.579951798437115]
  This work investigates iterative approximate evaluation for arbitrary prompts.<n>It introduces Model Predictive Prompt Selection (MoPPS), a Bayesian risk-predictive framework.<n>MoPPS reliably predicts prompt difficulty and accelerates training with significantly reduced rollouts.
  arXiv  Detail & Related papers  (2025-07-07T03:20:52Z)
• Beyond First-Order: Training LLMs with Stochastic Conjugate Subgradients and AdamW [2.028622227373579]
  gradient-based descent (SGD) have long been central to training large language models (LLMs)<n>This paper proposes a conjugate subgradient method together with adaptive sampling specifically for training LLMs.
  arXiv  Detail & Related papers  (2025-07-01T23:30:15Z)
• SPPD: Self-training with Process Preference Learning Using Dynamic Value Margin [16.346540681903804]
  We propose textbfSelf-training framework integrating textbfProcess textbfPreference learning using textbfDynamic value margin (SPPD)<n>Experiments on 7B-scale models demonstrate superior performance across in-domain and out-domain mathematical benchmarks.
  arXiv  Detail & Related papers  (2025-02-19T08:11:26Z)
• Bilevel ZOFO: Bridging Parameter-Efficient and Zeroth-Order Techniques for Efficient LLM Fine-Tuning and Meta-Training [44.48966200270378]
  Fine-tuning pre-trained Large Language Models (LLMs) for downstream tasks using First-Order (FO)imats presents significant computational challenges. We propose a bilevel optimization framework that complements ZO methods with PEFT to mitigate sensitivity to hard prompts. Our Bilevel ZOFO method employs a double-loop optimization strategy, where only the gradient of the PEFT model and the forward pass of the base model are required.
  arXiv  Detail & Related papers  (2025-02-05T20:47:44Z)
• Permutative Preference Alignment from Listwise Ranking of Human Judgments [40.23480751285947]
  We develop an end-to-end alignment algorithm by approximating NDCG with a differentiable surrogate loss.<n>We show that NDCG-based approaches improve ranking accuracy more effectively than B-T-based methods.
  arXiv  Detail & Related papers  (2024-10-06T03:49:28Z)
• Minor DPO reject penalty to increase training robustness [8.971332948872185]
  Learning from human preference is a paradigm used in large-scale language model (LLM) fine-tuning step to better align pretrained LLM to human preference for downstream task. Recently, Direct Preference Optimization (DPO) has been proposed to solve the alignment problem with a simplified RL-free method. In this article, we analyze the working mechanism of $beta$ in DPO, disclose its syntax difference between RL algorithm and DPO, and understand the potential shortage brought by the DPO simplification.
  arXiv  Detail & Related papers  (2024-08-19T09:29:31Z)
• SAIL: Self-Improving Efficient Online Alignment of Large Language Models [56.59644677997827]
  Reinforcement Learning from Human Feedback is a key method for aligning large language models with human preferences. Recent literature has focused on designing online RLHF methods but still lacks a unified conceptual formulation. Our approach significantly improves alignment performance on open-sourced datasets with minimal computational overhead.
  arXiv  Detail & Related papers  (2024-06-21T18:05:35Z)
• One-Shot Safety Alignment for Large Language Models via Optimal Dualization [64.52223677468861]
  This paper presents a perspective of dualization that reduces constrained alignment to an equivalent unconstrained alignment problem. We do so by pre-optimizing a smooth and convex dual function that has a closed form. Our strategy leads to two practical algorithms in model-based and preference-based settings.
  arXiv  Detail & Related papers  (2024-05-29T22:12:52Z)
• Decoding-time Realignment of Language Models [44.54462397717971]
  We propose a simple method to explore and evaluate different regularization strengths in aligned models without retraining. DeRa enables control over the degree of alignment, allowing users to smoothly transition between unaligned and aligned models.
  arXiv  Detail & Related papers  (2024-02-05T13:31:28Z)
• InferAligner: Inference-Time Alignment for Harmlessness through Cross-Model Guidance [56.184255657175335]
  We develop textbfInferAligner, a novel inference-time alignment method that utilizes cross-model guidance for harmlessness alignment. Experimental results show that our method can be very effectively applied to domain-specific models in finance, medicine, and mathematics. It significantly diminishes the Attack Success Rate (ASR) of both harmful instructions and jailbreak attacks, while maintaining almost unchanged performance in downstream tasks.
  arXiv  Detail & Related papers  (2024-01-20T10:41:03Z)
• Direct Preference Optimization: Your Language Model is Secretly a Reward Model [119.65409513119963]
  We introduce a new parameterization of the reward model in RLHF that enables extraction of the corresponding optimal policy in closed form. The resulting algorithm, which we call Direct Preference Optimization (DPO), is stable, performant, and computationally lightweight. Our experiments show that DPO can fine-tune LMs to align with human preferences as well as or better than existing methods.
  arXiv  Detail & Related papers  (2023-05-29T17:57:46Z)
• Pre-training Is (Almost) All You Need: An Application to Commonsense Reasoning [61.32992639292889]
  Fine-tuning of pre-trained transformer models has become the standard approach for solving common NLP tasks. We introduce a new scoring method that casts a plausibility ranking task in a full-text format. We show that our method provides a much more stable training phase across random restarts.
  arXiv  Detail & Related papers  (2020-04-29T10:54:40Z)

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.