Hidden Biases in Conditioning Autoregressive Models

• URL: http://arxiv.org/abs/2604.07855v1
• Date: Thu, 09 Apr 2026 06:12:16 GMT
• Title: Hidden Biases in Conditioning Autoregressive Models
• Abstract summary: Large language and music models are increasingly used for constrained generation.<n>These systems often perform strikingly well, but the induced procedures are usually not exact conditioning of the underlying autoregressive model.<n>We formalize several exact inference tasks for autoregressive models.
• Score: 0.0
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Large language and music models are increasingly used for constrained generation: rhyming lines, fixed meter, inpainting or infilling, positional endings, and other global form requirements. These systems often perform strikingly well, but the induced procedures are usually not exact conditioning of the underlying autoregressive model. This creates a hidden inferential bias, distinct from the better-known notion of bias inherited from the training set: samples are distorted relative to the true constrained distribution, with no generic guarantee of complete coverage of the admissible solution space or of correct conditional probabilities over valid completions. We formalize several exact inference tasks for autoregressive models and prove corresponding hardness results. For succinctly represented autoregressive models whose next-token probabilities are computable in polynomial time, exact sentence-level maximum a posteriori (MAP) decoding is NP-hard. This hardness persists under unary and metrical constraints. On the sampling side, exact conditioned normalization is \#P-hard even for regular constraints such as fixed-length terminal events. Unlike finite-state Markov models, general autoregressive models do not admit a bounded-state dynamic program for these tasks. These results formalize a standard claim in the neural decoding literature: local autoregressive sampling is easy, whereas exact decoding and exact conditioning under global form constraints are computationally intractable in general.

Related papers

• Constrained Decoding for Diffusion Language Models via Efficient Inference over Finite Automata [57.27430779838529]
  We present an exact and tractable algorithm for sampling from the constrained mean-field posterior under any constraint expressible as a finite automaton.<n>The approach guarantees constraint satisfaction by construction, supports both greedy and sampling-based decoding, and is compatible with parallel and block-wise decoding.<n> Empirical evaluations on Dream-7B and LLaDA-8 show substantial accuracy gains across various tasks.
  arXiv  Detail & Related papers  (2026-07-08T05:48:57Z)
• Sequential Kernel-based Conditional Independence Testing via Adaptive Betting [17.48164930490326]
  "Model-X" assumes exact knowledge of a relevant conditional distribution.<n>Existing sequential conditional independence tests require the Model-X conditional to be known exactly.<n>We propose a new approach that is substantially more robust to such estimation error.
  arXiv  Detail & Related papers  (2026-06-17T12:16:37Z)
• Adaptive Conformal Prediction for Improving Factuality of Generations by Large Language Models [86.8650252164764]
  Large language models (LLMs) are prone to generating factually incorrect outputs.<n>We propose an adaptive conformal prediction approach that extends conformal score transformation methods to LLMs.<n>This enables prompt-dependent calibration, retaining marginal coverage guarantees while improving conditional coverage.
  arXiv  Detail & Related papers  (2026-04-15T15:35:42Z)
• Continuous Diffusion Models Can Obey Formal Syntax [37.090659119764176]
  Diffusion language models offer a promising alternative to autoregressive models due to their global, non-causal process.<n>We introduce a training-free method for steering continuous language models to satisfy formal syntactic constraints expressed using regular expressions.<n>Our approach constructs an analytic score estimating the probability that a latent state decodes to a valid string accepted by a given regular expression.
  arXiv  Detail & Related papers  (2026-02-12T22:55:05Z)
• Can Continuous-Time Diffusion Models Generate and Solve Globally Constrained Discrete Problems? A Study on Sudoku [0.0]
  We use completed Sudoku grids as a controlled testbed, treating them as a subset of a continuous relaxation space.<n>We show that score-based samplers are the most reliable among continuous-time methods, and DDPM-style ancestral sampling achieves the highest validity overall.
  arXiv  Detail & Related papers  (2026-01-28T08:26:54Z)
• Fast Controlled Generation from Language Models with Adaptive Weighted Rejection Sampling [90.86991492288487]
  evaluating constraint on every token can be prohibitively expensive.<n> LCD can distort the global distribution over strings, sampling tokens based only on local information.<n>We show that our approach is superior to state-of-the-art baselines.
  arXiv  Detail & Related papers  (2025-04-07T18:30:18Z)
• It's All in the Mix: Wasserstein Classification and Regression with Mixed Features [2.2685251390114565]
  We develop and analyze distributionally robust prediction models that faithfully account for the presence of discrete features.<n>We demonstrate that our models can significantly outperform existing methods that are agnostic to the presence of discrete features.
  arXiv  Detail & Related papers  (2023-12-19T15:15:52Z)
• A Pseudo-Semantic Loss for Autoregressive Models with Logical Constraints [87.08677547257733]
  Neuro-symbolic AI bridges the gap between purely symbolic and neural approaches to learning. We show how to maximize the likelihood of a symbolic constraint w.r.t the neural network's output distribution. We also evaluate our approach on Sudoku and shortest-path prediction cast as autoregressive generation.
  arXiv  Detail & Related papers  (2023-12-06T20:58:07Z)
• User-defined Event Sampling and Uncertainty Quantification in Diffusion Models for Physical Dynamical Systems [49.75149094527068]
  We show that diffusion models can be adapted to make predictions and provide uncertainty quantification for chaotic dynamical systems. We develop a probabilistic approximation scheme for the conditional score function which converges to the true distribution as the noise level decreases. We are able to sample conditionally on nonlinear userdefined events at inference time, and matches data statistics even when sampling from the tails of the distribution.
  arXiv  Detail & Related papers  (2023-06-13T03:42:03Z)
• The Decaying Missing-at-Random Framework: Model Doubly Robust Causal Inference with Partially Labeled Data [8.916614661563893]
  We introduce a missing-at-random (decaying MAR) framework and associated approaches for doubly robust causal inference.<n>This simultaneously addresses selection bias in the labeling mechanism and the extreme imbalance between labeled and unlabeled groups.<n>To ensure robust causal conclusions, we propose a bias-reduced SS estimator for the average treatment effect.
  arXiv  Detail & Related papers  (2023-05-22T07:37:12Z)
• Generalised Bayesian Inference for Discrete Intractable Likelihood [9.331721990371769]
  This paper develops a novel generalised Bayesian inference procedure suitable for discrete intractable likelihood. Inspired by recent methodological advances for continuous data, the main idea is to update beliefs about model parameters using a discrete Fisher divergence. The result is a generalised posterior that can be sampled from using standard computational tools, such as Markov Monte Carlo.
  arXiv  Detail & Related papers  (2022-06-16T19:36:17Z)
• Consistency of a Recurrent Language Model With Respect to Incomplete Decoding [67.54760086239514]
  We study the issue of receiving infinite-length sequences from a recurrent language model. We propose two remedies which address inconsistency: consistent variants of top-k and nucleus sampling, and a self-terminating recurrent language model.
  arXiv  Detail & Related papers  (2020-02-06T19:56:15Z)

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.