Re-masking committed refusal tokens plus compliance prefixes bypasses safety in diffusion language models at 74-98% success across tested models.
hub Mixed citations
LLaDA 1.5: Variance-Reduced Preference Optimization for Large Language Diffusion Models
Mixed citation behavior. Most common role is background (62%).
63
Pith papers citing it
Background
62% of classified citations
abstract
While Masked Diffusion Models (MDMs), such as LLaDA, present a promising paradigm for language modeling, there has been relatively little effort in aligning these models with human preferences via reinforcement learning. The challenge primarily arises from the high variance in Evidence Lower Bound (ELBO)-based likelihood estimates required for preference optimization. To address this issue, we propose Variance-Reduced Preference Optimization (VRPO), a framework that formally analyzes the variance of ELBO estimators and derives bounds on both the bias and variance of preference optimization gradients. Building on this theoretical foundation, we introduce unbiased variance reduction strategies, including optimal Monte Carlo budget allocation and antithetic sampling, that significantly improve the performance of MDM alignment. We demonstrate the effectiveness of VRPO by applying it to LLaDA, and the resulting model, LLaDA 1.5, outperforms its SFT-only predecessor consistently and significantly across mathematical (GSM8K +4.7), code (HumanEval +3.0, MBPP +1.8), and alignment benchmarks (IFEval +4.0, Arena-Hard +4.3). Furthermore, LLaDA 1.5 demonstrates a highly competitive mathematical performance compared to strong language MDMs and ARMs. Project page: https://ml-gsai.github.io/LLaDA-1.5-Demo/.
hub tools
citation-role summary
background 5
dataset 1
method 1
other 1
citation-polarity summary
representative citing papers
Flow models reach 99.2% Sudoku accuracy in 7 passes and 96.1% on out-of-distribution Sudoku-Extreme by selecting dynamically stable candidates and training with self-conditioning plus DPO to avoid failed outputs.
Block-size curriculum learning trains an 8B diffusion model to achieve competitive reasoning performance on math and code benchmarks by transitioning from small to large training block sizes.
d-OPSD reframes on-policy self-distillation for dLLMs via suffix conditioning from self-generated answers and step-level supervision, outperforming RLVR and SFT on reasoning benchmarks with ~10% of the optimization steps.
D^2SD uses two diffusion drafters in a prefix tree structure with confidence scores to select and recover alternative draft sequences, achieving higher acceptance rates in speculative decoding.
MaskForge reaches 79.3% average attack success rate on five dLLMs by adaptively searching and accumulating structural attack patterns with a UCB bandit, improving 17.6% over baselines and transferring to 88.2% on AdvBench.
LIFT is a learnability-informed SFT algorithm for diffusion LMs that aligns token difficulty with diffusion time steps, yielding up to 3x gains on AIME'24 and AIME'25 over standard SFT baselines.
TokenDrift refines discrete diffusion language models by applying anti-symmetric drifting to soft-token features during training, yielding large reductions in generation perplexity at low NFEs.
PSD is a training-free framework that jointly optimizes spatial unmasking and temporal speculative decoding in diffusion LLMs to reach up to 5.5x tokens per forward pass while preserving accuracy comparable to greedy decoding.
TraFL applies trajectory flow balancing to post-train diffusion language models, preventing mode collapse and delivering consistent gains on reasoning tasks that hold under increased sampling.
Introduces Block-R1 benchmark, Block-R1-41K dataset, and a conflict score to handle domain-specific optimal block sizes in RL post-training of diffusion LLMs.
Infinite Mask Diffusion Models use stochastic infinite-state masks to overcome the factorization error lower bound in standard masked diffusion, achieving superior few-step performance on language tasks via distillation.
RSPO interprets reward advantages as targets for relative log-ratios in dLLMs, calibrating noisy estimates to stabilize RLVR training and achieve strong gains on planning tasks with competitive math reasoning performance.
TAD improves the accuracy-parallelism trade-off in diffusion LLMs via temporal-aware self-distillation that applies hard labels to soon-to-be-decoded tokens and soft supervision to future tokens.
ΔLPS is a gradient-guided discrete posterior sampler for inverse problems that works with masked or uniform discrete diffusion priors and outperforms prior discrete methods on image restoration tasks.
DualGaze-VLM uses text guidance and a new object-level dataset G-W3DA to predict driver attention, beating prior models by up to 17.8% in similarity metrics and passing human visual Turing tests at 88%.
Discrete Tilt Matching recasts dLLM fine-tuning as state-level matching of tilted local unmasking posteriors, producing a stable weighted cross-entropy loss that improves Sudoku and Countdown performance when applied to LLaDA-8B-Instruct.
NI Sampling accelerates discrete diffusion language models up to 14.3 times by training a neural indicator to select which tokens to sample at each step using a trajectory-preserving objective.
DepCap accelerates diffusion LM inference up to 5.63x by using last-block influence for adaptive block boundaries and conflict-free token selection for parallel decoding, with negligible quality loss.
DMax uses On-Policy Uniform Training and Soft Parallel Decoding to enable aggressive parallelism in dLLMs, raising TPF on GSM8K from 2.04 to 5.47 and on MBPP from 2.71 to 5.86 while preserving accuracy.
Visual Para-Thinker is the first parallel reasoning framework for MLLMs that uses visual partitioning strategies, Pa-Attention, and LPRoPE to extend test-time scaling benefits to visual comprehension tasks.
DisCa replaces heuristic feature caching with a lightweight learnable neural predictor compatible with distillation, achieving 11.8× acceleration on video diffusion transformers with preserved generation quality.
dMLLM-TTS delivers up to 6x more efficient test-time scaling for diffusion MLLMs via O(N+T) hierarchical search and self-verified feedback, improving generation quality on GenEval across three models.
Derives exact guidance transition rates for discrete flow matching models that require only one model evaluation per sampling step and unify prior approximation-based methods.
citing papers explorer
-
Stability-Weighted Decoding for Diffusion Language Models
Stability-Weighted Decoding improves diffusion LLM accuracy by modulating token scores with temporal stability from KL divergence between prediction steps.