Manisha Mukherjee

Research focus

Multimodal Models (1)Reasoning & Chain-of-Thought (1)RLHF & Preference Learning (1)Code Generation & Program Synthesis (1)Recommendation & Information Retrieval (1)

Frequent co-authors

Derek Lilienthal (1)Sameera Horawalavithana (1)Manisha Mukherjee (1)Vincent J. Hellendoorn (1)

Papers (2)

Apr 15, 2026

Apr 15, 2026·also JHU, National Laboratory, of the Rockies, RPI

Reward Design for Physical Reasoning in Vision-Language Models

Forget meticulously engineered reward functions – sometimes the simplest accuracy reward is all you need to make VLMs better at physics.

Derek Lilienthal, Manisha Mukherjee, Sameera Horawalavithana

Multimodal Models Reasoning & Chain-of-Thought RLHF & Preference Learning

Mar 2, 2026

Manisha Mukherjee +3Mar 2, 2026·also PNNL

Inference-Time Safety For Code LLMs Via Retrieval-Augmented Revision

Forget retraining: this retrieval-augmented method dynamically patches code LLMs against new vulnerabilities using Stack Overflow discussions, boosting security without model updates.

Manisha Mukherjee, Manisha Mukherjee, Vincent J. Hellendoorn +1

Code Generation & Program Synthesis Recommendation & Information Retrieval Red-Teaming & Adversarial Robustness

Search

Manisha Mukherjee

Research focus

Frequent co-authors

Papers (2)