6 papers
Ex-GraphRAG: Interpretable Evidence Routing for Graph-Augmented LLMs
Yoav Kor Sade, Arvindh Arun, Rishi Puri +2
GraphRAG conditions language models on subgraphs retrieved from knowledge graphs, encoded via message-passing GNNs. Because these encoders entangle node contributions through itera…
FutureSim: Replaying World Events to Evaluate Adaptive Agents
Shashwat Goel, Nikhil Chandak, Arvindh Arun +5
AI agents are being increasingly deployed in dynamic, open-ended environments that require adapting to new information as it arrives. To efficiently measure this capability for rea…
The Illusion of Diminishing Returns: Measuring Long Horizon Execution in LLMs
Akshit Sinha, Arvindh Arun, Shashwat Goel +2
Does continued scaling of large language models (LLMs) yield diminishing returns? In this work, we show that short-task benchmarks may give an illusion of slowing progress, as even…
SEMMA: A Semantic Aware Knowledge Graph Foundation Model
Arvindh Arun, Sumit Kumar, Mojtaba Nayyeri +4
Knowledge Graph Foundation Models (KGFMs) have shown promise in enabling zero-shot reasoning over unseen graphs by learning transferable patterns. However, most existing KGFMs rely…
A Cognac Shot To Forget Bad Memories: Corrective Unlearning for Graph Neural Networks
Varshita Kolipaka, Akshit Sinha, Debangan Mishra +4
Graph Neural Networks (GNNs) are increasingly being used for a variety of ML applications on graph data. Because graph data does not follow the independently and identically distri…
Topo Goes Political: TDA-Based Controversy Detection in Imbalanced Reddit Political Data
Arvindh Arun, Karuna K Chandra, Akshit Sinha +4
The detection of controversial content in political discussions on the Internet is a critical challenge in maintaining healthy digital discourse. Unlike much of the existing litera…