Minsu Kim

Generative Recursive Reasoning

Mingyu Jo

Mengye Ren

How should future neural reasoning systems implement extended computation? Recursive Reasoning Models (RRMs) offer a promising alternative t… (voir plus)o autoregressive sequence extension by performing iterative latent-state refinement with shared transition functions. Yet existing RRMs are largely deterministic, following a single latent trajectory and converging to a single prediction. We introduce Generative Recursive reAsoning Models (GRAM), a framework that turns recursive latent reasoning into probabilistic multi-trajectory computation. GRAM models reasoning as a stochastic latent trajectory, enabling multiple hypotheses, alternative solution strategies, and inference-time scaling through both recursive depth and parallel trajectory sampling. This yields a latent-variable generative model supporting conditional reasoning via

2026-05-18

arXiv (prépublication)

doi.org

arxiv.org

Synthesizable Molecular Generation via Soft-constrained GFlowNets with Rich Chemical Priors

Louis Vaillancourt

Yves V. Brun

Yoshua Bengio

Alex Hernández-García

2026-04-29

International Conference on Machine Learning (accepté)

doi.org

proceedings.mlr.press

Generative Recursive Reasoning Models

Mingyu Jo

We introduce Generative Recursive reAsoning Models (GRAM), a recursion-based generative model that is effective for complex planning and rea… (voir plus)soning problems. GRAM reformulates recent latent recursive architectures as a stochastic generative process with probabilistic latent transitions, enabling efficient and stable computation entirely in latent space without relying on token-level sequences as in chain-of-thought (CoT) prompting. We optimize this generative recursion via amortized variational inference, allowing the model to represent and explore multiple plausible latent trajectories conditioned on the input. This formulation supports both conditional reasoning through

2026-03-04

RSI @ International Conference on Learning Representations (poster)

openreview.net

Latent Veracity Inference for Identifying Errors in Stepwise Reasoning

Oliver E. Richardson

Sungsoo Ahn

Chain-of-Thought (CoT) reasoning has advanced the capabilities and transparency of language models (LMs); however, reasoning chains can cont… (voir plus)ain inaccurate statements that reduce performance and trustworthiness. To address this, we propose to augment each reasoning step in a CoT with a latent veracity (or correctness) variable. To efficiently explore this expanded space, we introduce Veracity Search (VS), a discrete search algorithm over veracity assignments. It performs otherwise intractable inference in the posterior distribution over latent veracity values by leveraging the LM's joint likelihood over veracity and the final answer as a proxy reward. This efficient inference-time verification method facilitates supervised fine-tuning of an Amortized Veracity Inference (AVI) machine by providing pseudo-labels for veracity. AVI generalizes VS, enabling accurate zero-shot veracity inference in novel contexts. Empirical results demonstrate that VS reliably identifies errors in logical (ProntoQA), mathematical (GSM8K), and commonsense (CommonsenseQA) reasoning benchmarks, with AVI achieving comparable zero-shot accuracy. Finally, we demonstrate the utility of latent veracity inference for providing feedback during self-correction and self-improvement.

2026-01-25

International Conference on Learning Representations (poster)

doi.org

openreview.net

Diffusion Alignment as Variataional Expectation-Maximization

Inhyuck Song

Hyeongyu Kang

Woocheol Shin

Taeyoung YUN

Kiyoung Om

Jinkyoo Park

Diffusion alignment aims to optimize diffusion models for the downstream objective. While existing methods based on reinforcement learning o… (voir plus)r direct backpropagation achieve considerable success in maximizing rewards, they often suffer from reward over-optimization and mode collapse. We introduce Diffusion Alignment as Variational Expectation-Maximization (DAV), a framework that formulates diffusion alignment as an iterative process alternating between two complementary phases: the E-step and the M-step. In the E-step, we employ test-time search to generate diverse and reward-aligned samples. In the M-step, we refine the diffusion model using samples discovered by the E-step. We demonstrate that DAV can optimize reward while preserving diversity for both continuous and discrete tasks: text-to-image synthesis and DNA sequence design.

2025-12-31

International Conference on Learning Representations (Accept (Poster))

openreview.net

Adaptive Inference-Time Scaling via Cyclic Diffusion Search

Gyubin Lee

Truong Nhat Nguyen Bao

Jaesik Yoon

Dongwoo Lee

Minsu Kim

Yoshua Bengio

Sungjin Ahn

Diffusion models have demonstrated strong generative capabilities across domains ranging from image synthesis to complex reasoning tasks. Ho… (voir plus)wever, most inference-time scaling methods rely on fixed denoising schedules, limiting their ability to allocate computation based on instance difficulty or task-specific demands adaptively. We introduce the challenge of adaptive inference-time scaling-dynamically adjusting computational effort during inference-and propose Adaptive Bi-directional Cyclic Diffusion (ABCD), a flexible, search-based inference framework. ABCD refines outputs through bi-directional diffusion cycles while adaptively controlling exploration depth and termination. It comprises three components: Cyclic Diffusion Search, Automatic Exploration-Exploitation Balancing, and Adaptive Thinking Time. Experiments show that ABCD improves performance across diverse tasks while maintaining computational efficiency.

2025-12-02

Conference on Neural Information Processing Systems (Accept (poster))

doi.org

openreview.net

Energy-based generator matching: A neural sampler for general state space

Dongyeop Woo

Minsu Kim

Minkyu Kim

Kiyoung Seong

Sungsoo Ahn

We propose Energy-based generator matching (EGM), a modality-agnostic approach to train generative models from energy functions in the absen… (voir plus)ce of data. Extending the recently proposed generator matching, EGM enables training of arbitrary continuous-time Markov processes, e.g., diffusion, flow, and jump, and can generate data from continuous, discrete, and a mixture of two modalities. To this end, we propose estimating the generator matching loss using self-normalized importance sampling with an additional bootstrapping trick to reduce variance in the importance weight. We validate EGM on both discrete and multimodal tasks up to 100 and 20 dimensions, respectively.

2025-12-02

Neural Information Processing Systems (Accept (poster))

doi.org

openreview.net

Trajectory Balance with Asynchrony: Decoupling Exploration and Learning for Fast, Scalable LLM Post-Training

Brian Bartoldson

James Diffenderfer

Tal Ben-Nun

Johan Obando-Ceron

Bhavya Kailkhura

Reinforcement learning (RL) is a critical component of large language model (LLM) post-training. However, on-policy algorithms used for post… (voir plus)-training are not naturally robust to a diversified content of experience replay buffers, which asynchronous off-policy actors can efficiently populate in parallel to training. We propose efficiently learning on such off-policy data via Trajectory Balance with Asynchrony (TBA), an approach to asynchronous RL for LLMs that leverages the principled off-policy TB objective. On math, preference-tuning, and automated red-teaming tasks, we post-train models ranging from Pythia 410M to Qwen 2.5 7B, finding TBA offers speed and performance boosts over strong baselines like Online DPO and Dr. GRPO. Beyond TBA's performance benefits (high accuracy even as asynchrony grows) and speedups (

2025-12-02

Neural Information Processing Systems (Accept (poster))

doi.org

openreview.net

Improved Off-policy Reinforcement Learning in Biological Sequence Design

Alex Hernández-García

Jinkyoo Park

Designing biological sequences with desired properties is challenging due to vast search spaces and limited evaluation budgets. Although rei… (voir plus)nforcement learning methods use proxy models for rapid reward evaluation, insufficient training data can cause proxy misspecification on out-of-distribution inputs. To address this, we propose a novel off-policy search,

2025-10-05

Proceedings of the 42nd International Conference on Machine Learning (publié)

doi.org

proceedings.mlr.press

Active Attacks: Red-teaming LLMs via Adaptive Environments

Taeyoung YUN

Pierre-Luc St-Charles

Jinkyoo Park

Yoshua Bengio

Minsu Kim

2025-09-25

ArXiv (prépublication)

doi.org

arxiv.org

Self-Evolving Curriculum for LLM Reasoning

Nicolas Gontier

Ehsan Kamalloo

2025-05-19

ArXiv (prépublication)

doi.org

arxiv.org

Ant Colony Sampling with GFlowNets for Combinatorial Optimization

Jiwoo Son

Jinkyoo Park

We present the Generative Flow Ant Colony Sampler (GFACS), a novel meta-heuristic method that hierarchically combines amortized inference an… (voir plus)d parallel stochastic search. Our method first leverages Generative Flow Networks (GFlowNets) to amortize a \emph{multi-modal} prior distribution over combinatorial solution space that encompasses both high-reward and diversified solutions. This prior is iteratively updated via parallel stochastic search in the spirit of Ant Colony Optimization (ACO), leading to the posterior distribution that generates near-optimal solutions. Extensive experiments across seven combinatorial optimization problems demonstrate GFACS's promising performances.

2025-05-02

International Conference on Artificial Intelligence and Statistics (Accept (Poster))

doi.org

proceedings.mlr.press

Formation d'entrepreneurs scientifiques

Communauté de pratique de Mila : L'IA agentique

Demandes de supervision

Publications

Formation d'entrepreneurs scientifiques

Communauté de pratique de Mila : L'IA agentique

Demandes de supervision

Mots-clés populaires:

Minsu Kim

Publications