arXiv subjects

Vishwas Sathish

Publications and source records attributed to Vishwas Sathish.

6 recordsLinked to original sources

arXiv2026

Eliciting Self-Verification in Multimodal Reasoning Agents with Reinforcement Learning

Reasoning agents increasingly rely on external tools such as web search to answer complex queries. Reinforcement learning (RL) finetuning algorithms such as GRPO have improved long-form reasoning in text-only language models, particularly for coding and mathematics. Reliable tool use in multimodal agents, however, remains challenging because models must interpret text and images while integrating noisy retrieved evidence, often under sparse outcome-level supervision without explicit verification signals. We present Self-Verification via Reinforcement Learning (SVRL), an RL-only finetuning framework that trains multimodal agents to verify and filter retrieved evidence within their own reasoning traces, reducing reliance on external verifiers at inference time. SVRL also introduces a search-aware penalty that discourages unnecessary tool calls and a query-diversity reward that encourages diverse, well-formed search queries, providing fine-grained feedback on when and what to search. Finetuning Qwen-2.5-VL-7B with SVRL on only 5{,}000 visual question answering examples yields consistent gains in multi-hop VQA generalization and tool efficiency across benchmarks. Overall, SVRL narrows the gap between compact agents and much larger proprietary models while requiring substantially lower training and inference cost.

cs.AI↗

arXiv2025

Lessons from Neuroscience for AI: How integrating Actions, Compositional Structure and Episodic Memory could enable Safe, Interpretable and Human-Like AI

The phenomenal advances in large language models (LLMs) and other foundation models over the past few years have been based on optimizing large-scale transformer models on the surprisingly simple objective of minimizing next-token prediction loss, a form of predictive coding that is also the backbone of an increasingly popular model of brain function in neuroscience and cognitive science. However, current foundation models ignore three other important components of state-of-the-art predictive coding models: tight integration of actions with generative models, hierarchical compositional structure, and episodic memory. We propose that to achieve safe, interpretable, energy-efficient, and human-like AI, foundation models should integrate actions, at multiple scales of abstraction, with a compositional generative architecture and episodic memory. We present recent evidence from neuroscience and cognitive science on the importance of each of these components. We describe how the addition of these missing components to foundation models could help address some of their current deficiencies: hallucinations and superficial understanding of concepts due to lack of grounding, a missing sense of agency/responsibility due to lack of control, threats to safety and trustworthiness due to lack of interpretability, and energy inefficiency. We compare our proposal to current trends, such as adding chain-of-thought (CoT) reasoning and retrieval-augmented generation (RAG) to foundation models, and discuss new ways of augmenting these models with brain-inspired components. We conclude by arguing that a rekindling of the historically fruitful exchange of ideas between brain science and AI will help pave the way towards safe and interpretable human-centered AI.

cs.AI↗

arXiv2025

PainDECOG: Machine Learning-Based Identification of Pain Biomarkers from sEEG Signals

This study presents a systematic machine-learning approach for classifying acute pain from raw electrophysiological signals. We address binary and ternary classification tasks, leveraging Power-In-Band (PIB) and signal coherence as distinguishing features. Our method evaluates the effectiveness of traditional machine learning algorithms on a manually curated electrophysiological dataset obtained from intracranial electroencephalography (iEEG), offering valuable insights into model performance for pain detection. Furthermore, we identify critical electrode pairings associated with acute pain, providing a clearer understanding of the neural markers that differentiate pain states. This work highlights the potential of targeted feature engineering in advancing pain classification, setting the stage for future enhancements in real-time and personalized pain assessment tools. Additionally, these findings have promising applications in neuromodulation and Deep Brain Stimulation (DBS), where adaptive and closed-loop systems could leverage identified pain markers to modulate pain-related brain regions more precisely, offering improved therapeutic options for chronic pain management

eess.SP↗

arXiv2024

LLeMpower: Understanding Disparities in the Control and Access of Large Language Models

Large Language Models (LLMs) are a powerful technology that augment human skill to create new opportunities, akin to the development of steam engines and the internet. However, LLMs come with a high cost. They require significant computing resources and energy to train and serve. Inequity in their control and access has led to concentration of ownership and power to a small collection of corporations. In our study, we collect training and inference requirements for various LLMs. We then analyze the economic strengths of nations and organizations in the context of developing and serving these models. Additionally, we also look at whether individuals around the world can access and use this emerging technology. We compare and contrast these groups to show that these technologies are monopolized by a surprisingly few entities. We conclude with a qualitative study on the ethical implications of our findings and discuss future directions towards equity in LLM access.

cs.CY↗

arXiv2022

Active Predictive Coding: A Unified Neural Framework for Learning Hierarchical World Models for Perception and Planning

Predictive coding has emerged as a prominent model of how the brain learns through predictions, anticipating the importance accorded to predictive learning in recent AI architectures such as transformers. Here we propose a new framework for predictive coding called active predictive coding which can learn hierarchical world models and solve two radically different open problems in AI: (1) how do we learn compositional representations, e.g., part-whole hierarchies, for equivariant vision? and (2) how do we solve large-scale planning problems, which are hard for traditional reinforcement learning, by composing complex action sequences from primitive policies? Our approach exploits hypernetworks, self-supervised learning and reinforcement learning to learn hierarchical world models that combine task-invariant state transition networks and task-dependent policy networks at multiple abstraction levels. We demonstrate the viability of our approach on a variety of vision datasets (MNIST, FashionMNIST, Omniglot) as well as on a scalable hierarchical planning problem. Our results represent, to our knowledge, the first demonstration of a unified solution to the part-whole learning problem posed by Hinton, the nested reference frames problem posed by Hawkins, and the integrated state-action hierarchy learning problem in reinforcement learning.

cs.LG↗

arXiv2019

Graph Embedding Based Hybrid Social Recommendation System

Item recommendation tasks are a widely studied topic. Recent developments in deep learning and spectral methods paved a path towards efficient graph embedding techniques. But little research has been done on applying these graph embedding to social graphs for recommendation tasks. This paper focuses at performance of various embedding methods applied on social graphs for the task of item recommendation. Additionally, a hybrid model is proposed wherein chosen embedding models are combined together to give a collective output. We put forward the hypothesis that such a hybrid model would perform better than individual embedding for recommendation task. With recommendation using individual embedding as a baseline, performance for hybrid model for the same task is evaluated and compared. Standard metrics are used for qualitative comparison. It is found that the proposed hybrid model outperforms the baseline.

cs.SI↗