Overview

Language model decoding is the process that turns a model’s next-token predictions into text, and its choices affect accuracy, factuality, style, and speed. Common approaches include greedy, temperature, top-k and nucleus sampling, beam search, and best-of-many selection; newer methods use uncertainty, reward models, verifiers, or retrieved context to guide choices during generation. Work also explores parallel and diffusion-based decoding, speculative generation, and attention or cache optimizations to reduce latency and memory use. A central challenge is balancing quality and reliability against cost and diversity, since confidence and reward signals do not always identify the best continuation and gains can vary by task and evaluation setting.

Latest papers

182 papers19 papers in the last four weeks, down 27% on the four weeks before. 0.1% of all new papers.

Mar 4, 2025

Evolutionary Guided Decoding: Iterative Value Refinement for LLMs

While guided decoding, especially value-guided methods, has emerged as a cost-effective alternative for controlling language model outputs without re-training models, its effectiveness is limited by the accuracy of the value function. We identify that this inaccuracy stems from a core distributional gap: existing methods train static value functions on trajectories sampled exclusively from the base policy, which inherently confines their training to a narrow and suboptimal view of the potential output space. We propose Iterative Value Refinement, a evolutionary framework designed to narrow this gap. It employs Value Exploration to provide a more comprehensive and robust training signal, complemented by Iterative Self-Refinement, which uses the improved value function from one iteration to guide the generation of higher-quality data for the next. Extensive experiments on text summarization, multi-turn dialogue, and instruction following demonstrate the effectiveness of our framework in aligning language models. Our approach not only achieves alignment but also significantly reduces computational costs by leveraging principled value function optimization for efficient and effective control.
Dec 10, 2024

DYNASHIELD: A Black-Box Moving Target Defense for LLMs via Dynamic Decoding Customization

Large language models (LLMs) remain vulnerable to jailbreak attacks in which adversarial prompts induce harmful outputs. Existing defenses often require access to the model internals or additional training, limiting their applicability for service providers deployed through black-box APIs. In this paper, we propose DYNASHIELD, a moving target defense framework that improves robustness by customizing decoding hyperparameters and system prompts at inference time. DYNASHIELD includes two key steps: (1) it identifies decoding configurations that reduce attack success probability, and (2) it probabilistically samples from a weighted configuration pool to introduce controlled variability in model behavior. We evaluate DYNASHIELD across 7 open-source LLMs under 4 state-of-the-art jailbreak attacks, using adversarial prompts from AdvBench. Results show substantial reductions in attack success rate compared with 7 baseline defenses, while maintaining response quality and incurring minimal inference overhead. Because DYNASHIELD operates solely through exposed runtime controls, it requires no retraining or model-internal access. These results suggest that safety-aware dynamic decoding is a promising and practically lightweight defense mechanism for black-box LLM deployments.