# Manan Suri > Manan Suri is a Computer Science PhD student at the University of Maryland, College Park (GAMMA Lab, advised by Dinesh Manocha). He works on LLM agents: how agents hold and use context, attribution methods that make agent outputs trustworthy, parametric memory that distills multimodal context into LoRA adapters, and the training data and benchmarks agents need. Not to be confused with Manan Suri, professor at IIT Delhi. Site: https://manansuri.com Contact: see https://manansuri.com/ (email, X @mananifesto, GitHub MananSuri27) Identifiers: ORCID https://orcid.org/0000-0001-5089-2658 · Google Scholar https://scholar.google.com/citations?user=KbVluf4AAAAJ · Semantic Scholar https://www.semanticscholar.org/author/145282060 ## Pages - [About](/): bio, news, press, talks, service - [Publications](/publications/): all papers with abstracts, PDF, code, and bibtex - Paper pages (one page per paper with abstract, figure, results, links, and bibtex): - [Omni2LoRA: Coherence-Preserving Parametric Memory for Efficient Omni Language Models](/papers/omni2lora/) (arXiv 2026) - [CyberForge: Verified Vulnerability Injection at Repository Level for Cybersecurity Agent Training](/papers/cyberforge/) (arXiv 2026) - [Frames2LoRA: Parametric Video Internalization for Vision-Language Models](/papers/frames2lora/) (arXiv 2026) - [CodeScout: Contextual Problem Statement Enhancement for Software Agents](/papers/codescout/) (ACL 2026 Findings 2026) - [Structured Uncertainty Guided Clarification for LLM Agents](/papers/sage-clarification/) (ACL 2026 Findings 2026) - [Follow the Flow: Fine-grained Flowchart Attribution with Neurosymbolic Agents](/papers/follow-the-flow/) (EMNLP 2025 2025) - [ChartLens: Fine-grained Visual Attribution in Charts](/papers/chartlens/) (ACL 2025 2025) - [VisDoM: Multi-Document QA with Visually Rich Elements Using Multimodal Retrieval-Augmented Generation](/papers/visdom/) (NAACL 2025 2025) - [DocEdit-v2: Document Structure Editing Via Multimodal LLM Grounding](/papers/docedit-v2/) (EMNLP 2024 2024) - [Doc2Command: Furthering Language Guided Document Editing](/papers/doc2command/) (ICLR 2024 2024) - [Trigger Warnings: A Computational Approach to Understanding User-Tagged Trigger Warnings](/papers/trigger-warnings/) (RANLP 2023 SRW 2023) - [WADER: A Weak-labelling framework for Data augmentation in tExt Regression Tasks](/papers/wader/) (SemEval 2023 2023) - [ACLM: A Selective-Denoising based Generative Data Augmentation Approach for Low-Resource Complex NER](/papers/aclm/) (ACL 2023 2023) - [The Geometry of Multilingual Language Models: A Fairness Lens](/papers/multilingual-geometry-fairness/) (ICLR 2023 Tiny Papers 2023) - [CoSyn: Detecting Implicit Hate Speech in Online Conversations Using a Context Synergized Hyperbolic Network](/papers/cosyn/) (EMNLP 2023 2023) - [Multimodal Analysis and Modality Fusion for Detection of Depression from Twitter Data](/papers/depression-twitter-multimodal/) (AI4SG @ AAAI 2023 2023) - [I Don't Feel so Good! Detecting Depressive Tendencies Using Transformer-Based Multimodal Frameworks](/papers/depression-transformer-multimodal/) (MLNLP 2022 2022) - [Multilingual Protest Event Detection using Transformer-based Models](/papers/protest-event-detection/) (CASE 2022 2022) - [A novel multimodal dynamic fusion network for disfluency detection in spoken utterances](/papers/disfluency-detection/) (arXiv 2022 2022) - [Multimodal BERT for Identifying Claims in Tweets](/papers/claim-detection-tweets/) (CLEF 2022 2022) - [Boosting Pre-trained Language Models with Task Specific Metadata and Cost Sensitive Learning](/papers/pcl-detection/) (SemEval-2022 2022) - [Blog](/blog/): explainers and notes - [News](/news/): dated updates - [CV](/cv/): PDF résumé - [Full text for LLMs](/llms-full.txt): everything on this site in one markdown file ## Selected papers - Frames2LoRA: Parametric Video Internalization for Vision-Language Models (2026) — https://arxiv.org/abs/2606.04351 - CyberForge: Verified Vulnerability Injection at Repository Level for Cybersecurity Agent Training (2026) — https://arxiv.org/abs/2608.06471 · project page https://cyb3rforge.github.io/ - Omni2LoRA: Coherence-Preserving Parametric Memory for Efficient Omni Language Models (2026) — https://arxiv.org/abs/2608.09227 - CodeScout: Contextual Problem Statement Enhancement for Software Agents (ACL 2026 Findings) — https://arxiv.org/abs/2603.05744 - Structured Uncertainty Guided Clarification for LLM Agents (ACL 2026 Findings) — https://arxiv.org/abs/2511.08798 - ChartLens: Fine-grained Visual Attribution in Charts (ACL 2025) — https://aclanthology.org/2025.acl-long.1094/ - Follow the Flow: Fine-grained Flowchart Attribution with Neurosymbolic Agents (EMNLP 2025) — https://aclanthology.org/2025.emnlp-main.1144/ - VisDoM: Multi-Document QA with Visually Rich Elements Using Multimodal RAG (NAACL 2025) — https://arxiv.org/abs/2412.10704