8 papers
Extrapolative Weight Averaging Reveals Correctness-Efficiency Frontiers in Code RL
Kunhao Zheng, Pierre Chambon, Juliette Decugis +4
Linear interpolation between fine-tuned checkpoints has been shown to trace the Pareto front between competing objectives, but whether extrapolative weight averaging can extend suc…
Beyond pass@k: Redundancy-Aware RLVR for Multi-Sample Code Generation
Le Bronnec Florian, Alexandre Verine, Rio Yokota +1
LLMs for code generation are commonly evaluated in repeated-sampling settings using Pass@k, where multiple candidate programs are executed against unit tests under a finite samplin…
Improving Diversity in Language Models: When Temperature Fails, Change the Loss
Alexandre Verine, Florian Le Bronnec, Kunhao Zheng +3
Increasing diversity in language models is a challenging yet essential objective. A common approach is to raise the decoding temperature. In this work, we investigate this approach…
Lattice Climber Attack: Adversarial attacks for randomized mixtures of classifiers
Lucas Gnecco-Heredia, Benjamin Negrevergne, Yann Chevaleyre
Finite mixtures of classifiers (a.k.a. randomized ensembles) have been proposed as a way to improve robustness against adversarial attacks. However, existing attacks have been show…
Unveiling the Role of Randomization in Multiclass Adversarial Classification: Insights from Graph Theory
Lucas Gnecco-Heredia, Matteo Sammut, Muni Sreenivas Pydi +3
Randomization as a mean to improve the adversarial robustness of machine learning models has recently attracted significant attention. Unfortunately, much of the theoretical analys…
Improving Discriminator Guidance in Diffusion Models
Alexandre Verine, Ahmed Mehdi Inane, Florian Le Bronnec +2
Discriminator Guidance has become a popular method for efficiently refining pre-trained Score-Matching Diffusion models. However, in this paper, we demonstrate that the standard im…