Abhinav Verma
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Abhinav Verma has a total of 9 co-authors. Here are the top co-authors by number of shared publications:
Jonathan Dodge
1 shared publication- Krishnendu Chatterjee
IST Austria, Klosterneuburg, Austria
1 shared publication - Mathias Lechner
Massachusetts Institute of Technology
1 shared publication Shreyash Kale
1 shared publicationSourav Panda
1 shared publicationTanmay Ambadkar
1 shared publication
Latest Publications
- Open Access
Preference Conditioned Multi-Objective Reinforcement Learning: Decomposed, Diversity-Driven Policy Optimization
Multi-objective reinforcement learning (MORL) seeks to learn policies that balance multiple, often conflicting objectives. Although a single preference-conditioned policy is the most flexible and scalable solution, existing approaches remain brittle in practice, frequently failing to recover complete Pareto fronts. We show that this failure stems from two structural issues in current methods: destructive gradient interference caused by premature scalarization and representational collapse across the preference space. We introduce $D^3PO$, a PPO-based framework that reorganizes multi-objective policy optimization to address these issues directly. $D^3PO$ preserves per-objective learning signals through a decomposed optimization pipeline and integrates preferences only after stabilization, enabling reliable credit assignment. In addition, a scaled diversity regularizer enforces sensitivity of policy behavior to preference changes, preventing collapse. Across standard MORL benchmarks, including high-dimensional and many-objective control tasks, $D^3PO$ consistently discovers broader and higher-quality Pareto fronts than prior single- and multi-policy methods, matching or exceeding state-of-the-art hypervolume and expected utility while using a single deployable policy.Open AccessPreference Conditioned Multi-Objective Reinforcement Learning: Decomposed, Diversity-Driven Policy Optimization
Multi-objective reinforcement learning (MORL) seeks to learn policies that balance multiple, often conflicting objectives. Although a single preference-conditioned policy is the most flexible and scalable solution, existing approaches remain brittle in practice, frequently failing to recover complete Pareto fronts. We show that this failure stems from two structural issues in current methods: destructive gradient interference caused by premature scalarization and representational collapse across the preference space. We introduce $D^3PO$, a PPO-based framework that reorganizes multi-objective policy optimization to address these issues directly. $D^3PO$ preserves per-objective learning signals through a decomposed optimization pipeline and integrates preferences only after stabilization, enabling reliable credit assignment. In addition, a scaled diversity regularizer enforces sensitivity of policy behavior to preference changes, preventing collapse. Across standard MORL benchmarks, including high-dimensional and many-objective control tasks, $D^3PO$ consistently discovers broader and higher-quality Pareto fronts than prior single- and multi-policy methods, matching or exceeding state-of-the-art hypervolume and expected utility while using a single deployable policy. - Open Access
- 1 September 2025
- Published by International Joint Conferences on Artificial Intelligence
- Vol. 9, 7509-7517
RetroMoE: A Mixture-of-Experts Latent Translation Framework for Single-step Retrosynthesis
Single-step retrosynthesis is a crucial task in organic synthesis, where the objective is to identify the reactants needed to produce a given product. In recent years, a variety of machine learning methods have been developed to tackle retrosynthesis prediction. In our study, we introduce RetroMoE, a novel generative model designed for the single-step retrosynthesis task. We start with a non-symmetric variational autoencoder (VAE) that incorporates a graph encoder to map molecular graphs into a latent space, followed by a transformer decoder for precise prediction of molecular SMILES strings. Additionally, we implement a simple yet effective mixture-of-experts (MoE) network to translate the product latent embedding into the reactant latent embedding. To our knowledge, this is the first approach that frames single-step retrosynthesis as a latent translation problem. Extensive experiments on the USPTO-50K and USPTO-MIT datasets demonstrate the superiority of our method, which not only surpasses most semi-template-based and template-free methods but also delivers competitive results against template-based methods. Notably, under the class-known setting on the USPTO-50K, our method achieves top-1 exact match accuracy comparable to the state-of-the-art template method, RetroKNN.
- 1 September 2025
- Published by International Joint Conferences on Artificial Intelligence
- Vol. 9, 7509-7517
Open AccessRetroMoE: A Mixture-of-Experts Latent Translation Framework for Single-step Retrosynthesis
Single-step retrosynthesis is a crucial task in organic synthesis, where the objective is to identify the reactants needed to produce a given product. In recent years, a variety of machine learning methods have been developed to tackle retrosynthesis prediction. In our study, we introduce RetroMoE, a novel generative model designed for the single-step retrosynthesis task. We start with a non-symmetric variational autoencoder (VAE) that incorporates a graph encoder to map molecular graphs into a latent space, followed by a transformer decoder for precise prediction of molecular SMILES strings. Additionally, we implement a simple yet effective mixture-of-experts (MoE) network to translate the product latent embedding into the reactant latent embedding. To our knowledge, this is the first approach that frames single-step retrosynthesis as a latent translation problem. Extensive experiments on the USPTO-50K and USPTO-MIT datasets demonstrate the superiority of our method, which not only surpasses most semi-template-based and template-free methods but also delivers competitive results against template-based methods. Notably, under the class-known setting on the USPTO-50K, our method achieves top-1 exact match accuracy comparable to the state-of-the-art template method, RetroKNN.
- Open Access
Compositional Policy Learning in Stochastic Control Systems with Formal Guarantees
Reinforcement learning has shown promising results in learning neural network policies for complicated control tasks. However, the lack of formal guarantees about the behavior of such policies remains an impediment to their deployment. We propose a novel method for learning a composition of neural network policies in stochastic environments, along with a formal certificate which guarantees that a specification over the policy's behavior is satisfied with the desired probability. Unlike prior work on verifiable RL, our approach leverages the compositional nature of logical specifications provided in SpectRL, to learn over graphs of probabilistic reach-avoid specifications. The formal guarantees are provided by learning neural network policies together with reach-avoid supermartingales (RASM) for the graph's sub-tasks and then composing them into a global policy. We also derive a tighter lower bound compared to previous work on the probability of reach-avoidance implied by a RASM, which is required to find a compositional policy with an acceptable probabilistic threshold for complex tasks with multiple edge policies. We implement a prototype of our approach and evaluate it on a Stochastic Nine Rooms environment.
Open AccessCompositional Policy Learning in Stochastic Control Systems with Formal Guarantees
Reinforcement learning has shown promising results in learning neural network policies for complicated control tasks. However, the lack of formal guarantees about the behavior of such policies remains an impediment to their deployment. We propose a novel method for learning a composition of neural network policies in stochastic environments, along with a formal certificate which guarantees that a specification over the policy's behavior is satisfied with the desired probability. Unlike prior work on verifiable RL, our approach leverages the compositional nature of logical specifications provided in SpectRL, to learn over graphs of probabilistic reach-avoid specifications. The formal guarantees are provided by learning neural network policies together with reach-avoid supermartingales (RASM) for the graph's sub-tasks and then composing them into a global policy. We also derive a tighter lower bound compared to previous work on the probability of reach-avoidance implied by a RASM, which is required to find a compositional policy with an acceptable probabilistic threshold for complex tasks with multiple edge policies. We implement a prototype of our approach and evaluate it on a Stochastic Nine Rooms environment.