Balancing Diversity and Risk in LLM Sampling: How to Select Your Method and Parameter for Open-Ended Text Generation

• URL: http://arxiv.org/abs/2408.13586v1
• Date: Sat, 24 Aug 2024 14:14:32 GMT
• Title: Balancing Diversity and Risk in LLM Sampling: How to Select Your Method and Parameter for Open-Ended Text Generation
• Authors: Yuxuan Zhou, Margret Keuper, Mario Fritz,
• Abstract summary: We propose a systematic way to estimate the intrinsic capacity of a truncation sampling method by considering the trade-off between diversity and risk at each decoding step. Our work provides a comprehensive comparison between existing truncation sampling methods, as well as their recommended parameters as a guideline for users.
• Score: 60.493180081319785
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Sampling-based decoding strategies have been widely adopted for Large Language Models (LLMs) in numerous applications, which target a balance between diversity and quality via temperature tuning and tail truncation (e.g., top-k and top-p sampling). Considering the high dynamic range of the candidate next-token given different prefixes, recent studies propose to adaptively truncate the tail of LLM's predicted distribution. Although improved results haven been reported with these methods on open-ended text generation tasks, the results are highly dependent on the curated truncation parameters and exemplar text. In this paper, we propose a systematic way to estimate the intrinsic capacity of a truncation sampling method by considering the trade-off between diversity and risk at each decoding step, based on our collected prefix tree which preserves the context of a full sentence. Our work provides a comprehensive comparison between existing truncation sampling methods, as well as their recommended parameters as a guideline for users.

Related papers

• PEDAL: Enhancing Greedy Decoding with Large Language Models using Diverse Exemplars [1.450405446885067]
  Self-ensembling techniques with diverse reasoning paths have demonstrated remarkable performance gains in text generation with Large Language Models (LLMs) We introduce PEDAL, a hybrid self-ensembling approach that combines the strengths of diverse exemplar based prompts and LLM based aggregation to achieve improvement in overall performance.
  arXiv  Detail & Related papers  (2024-08-16T17:54:09Z)
• Turning Up the Heat: Min-p Sampling for Creative and Coherent LLM Outputs [4.122612309805664]
  Large Language Models (LLMs) generate text by sampling the next token from a probability distribution over the vocabulary at each decoding step. We propose min-p sampling, a dynamic truncation method that adjusts the sampling threshold based on the model's confidence by scaling according to the top token's probability. We conduct extensive experiments on benchmarks including GPQA, GSM8K, and AlpacaEval Creative Writing, demonstrating that min-p sampling improves both the quality and diversity of generated text, particularly at high temperatures.
  arXiv  Detail & Related papers  (2024-07-01T08:37:25Z)
• Adaptive Online Bayesian Estimation of Frequency Distributions with Local Differential Privacy [0.4604003661048266]
  We propose a novel approach for the adaptive and online estimation of the frequency distribution of a finite number of categories under the local differential privacy (LDP) framework. The proposed algorithm performs Bayesian parameter estimation via posterior sampling and adapts the randomization mechanism for LDP based on the obtained posterior samples. We provide a theoretical analysis showing that (i) the posterior distribution targeted by the algorithm converges to the true parameter even for approximate posterior sampling, and (ii) the algorithm selects the optimal subset with high probability if posterior sampling is performed exactly.
  arXiv  Detail & Related papers  (2024-05-11T13:59:52Z)
• Amortizing intractable inference in large language models [56.92471123778389]
  We use amortized Bayesian inference to sample from intractable posterior distributions. We empirically demonstrate that this distribution-matching paradigm of LLM fine-tuning can serve as an effective alternative to maximum-likelihood training. As an important application, we interpret chain-of-thought reasoning as a latent variable modeling problem.
  arXiv  Detail & Related papers  (2023-10-06T16:36:08Z)
• Language Model Decoding as Direct Metrics Optimization [87.68281625776282]
  Current decoding methods struggle to generate texts that align with human texts across different aspects. In this work, we frame decoding from a language model as an optimization problem with the goal of strictly matching the expected performance with human texts. We prove that this induced distribution is guaranteed to improve the perplexity on human texts, which suggests a better approximation to the underlying distribution of human texts.
  arXiv  Detail & Related papers  (2023-10-02T09:35:27Z)
• Bilevel Scheduled Sampling for Dialogue Generation [6.89978591161039]
  We propose a bilevel scheduled sampling model that takes the sentence-level information into account and incorporates it with word-level quality. Experiments conducted on the DailyDialog and PersonaChat datasets demonstrate the effectiveness of our proposed methods.
  arXiv  Detail & Related papers  (2023-09-05T05:05:06Z)
• Finding Support Examples for In-Context Learning [73.90376920653507]
  We propose LENS, a fiLter-thEN-Search method to tackle this challenge in two stages. First we filter the dataset to obtain informative in-context examples individually. Then we propose diversity-guided example search which iteratively refines and evaluates the selected example permutations.
  arXiv  Detail & Related papers  (2023-02-27T06:32:45Z)
• Arithmetic Sampling: Parallel Diverse Decoding for Large Language Models [65.52639709094963]
  Methods such as beam search and Gumbel top-k sampling can guarantee a different output for each element of the beam, but are not easy to parallelize. We present a framework for sampling according to an arithmetic code book implicitly defined by a large language model.
  arXiv  Detail & Related papers  (2022-10-18T22:19:41Z)
• A Well-Composed Text is Half Done! Composition Sampling for Diverse Conditional Generation [79.98319703471596]
  We propose Composition Sampling, a simple but effective method to generate diverse outputs for conditional generation of higher quality. It builds on recently proposed plan-based neural generation models that are trained to first create a composition of the output and then generate by conditioning on it and the input.
  arXiv  Detail & Related papers  (2022-03-28T21:24:03Z)

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.