Amir Saeidi

I am a fourth year PhD student in CS at Arizona State University, advised by Chitta Baral in the Cognition and Intelligence Lab. I previously interned at Microsoft Research, where I worked on automating synthetic data generation pipelines to build capable LLM agents.

My research focuses on post-training and alignment of LLMs (TPO, DCPO), building reliable agentic systems (IRMA, FAMA), and scaling synthetic data generation for agent learning (VULCAN, CAST, ML-AutoResearch). More broadly, I am interested in developing reliable LLM agents for complex real-world settings, as well as agentic systems for scientific discovery and automated research.

I am actively looking for industry research positions focused on agentic systems, LLM post-training, and synthetic data generation. Feel free to reach out!

Amir Saeidi profile photo

News

  • September, 2026 - ML-AutoResearch accepted at NeurIPS 2026.
  • September, 2026 - VULCAN implementation is now publicly available on Microsoft's GitHub.
  • August, 2026 - CAST accepted at EMNLP 2026.
  • August, 2026 - ML-AutoResearch is out on arXiv. Work done during internship at Microsoft Research.
  • May, 2026 - LedgerAgent is out on arXiv.
  • May, 2026 - Returning to Microsoft Research as a Research Intern for Summer 2026.
  • May, 2026 - Contributed to Microsoft's MagenticLite and Aion 1.0 Instinct releases.
  • May, 2026 - Presented VULCAN at the MSR Forum.
  • April, 2026 - TPO added to HuggingFace TRL.
  • April, 2026 - FAMA accepted at ACL 2026 Findings.
  • March, 2026 - VULCAN accepted at DATA-FM @ ICLR 2026. Work done during internship at Microsoft Research.
  • November, 2025 - TPO accepted at TMLR.
  • October, 2025 - Continuing at Microsoft Research as a part-time Research Intern.
  • September, 2025 - 2 papers accepted at NeurIPS 2025 workshops.
  • September, 2025 - DCPO accepted at TMLR.
  • August, 2025 - IRMA accepted at EMNLP 2025 Findings.
  • May, 2025 - UnSeenTimeQA accepted at ACL 2025.
  • January, 2025 - Joining Microsoft Research as a Research Intern for Spring and Summer 2025.
  • April, 2024 - 1 paper accepted at CVPR 2024 workshop.
  • October, 2023 - Started my research fellowship at Mayo Clinic.
  • January, 2023 - Started my PhD at ASU.

Publications

ML-AutoResearch overview

ML-AutoResearch: Training Machine Learning Research Agents with Automatically Generated Environments

Ziyang Cai*, Amir Saeidi*, Harkirat Behl

NeurIPS 2026

Paper

CAST overview

CAST: Critique-Aware Supervision for Training Reliable Long-Horizon Tool-Calling Agents

Amir Saeidi*, Zehua Zhang*, Rishitosh Singh, Naman Ahuja, Vivek Gupta, Ali Payani, Gaowen Liu, Jayanth Srinivasa, Chitta Baral

EMNLP 2026

Paper

LedgerAgent overview

LedgerAgent: Structured State for Policy-Adherent Tool-Calling Agents

Md Nayem Uddin, Amir Saeidi, Eduardo Blanco, Chitta Baral

Under Review

Paper

VULCAN environment simulator overview VULCAN task configuration overview

VULCAN: Where Agents Learn by Living in Simulated Tool Environments

Amir Saeidi, Chitta Baral, Ahmed Awadallah, Harkirat Behl

DATA-FM Workshop at ICLR 2026

Paper/ Code

FAMA overview

FAMA: Failure-Aware Meta-Agentic Framework for Open-Source LLMs in Interactive Tool Use Environments

Amir Saeidi*, Venkatesh Mishra*, Souradeep Mukhopadhyay, Gaowen Liu, Ali Payani, Jayanth Srinivasa, Chitta Baral

ACL 2026 Findings

Paper

IRMA overview

How Can Input Reformulation Improve Tool Usage Accuracy in a Complex Dynamic Environment? A Study on ฯ„-bench

Venkatesh Mishra*, Amir Saeidi*, Satyam Raj, Mutsumi Nakamura, Jayanth Srinivasa, Gaowen Liu, Ali Payani, Chitta Baral

EMNLP 2025 Findings ยท MTI-LLM Workshop at NeurIPS 2025

Paper

TPO overview

Triple Preference Optimization: Achieving Better Alignment using a Single Step Optimization

Amir Saeidi, Shivanshu Verma, Kashif Rasul, Aswin RRV, Chitta Baral

TMLR

Paper/ Project/ Code/ TRL

DCPO overview

Dual Caption Preference Optimization for Diffusion Models

Amir Saeidi*, Yiran Luo*, Agneet Chatterjee, Shamanthak Hegde, Bimsara Pathiraja, Yezhou Yang, Chitta Baral

TMLR

Paper/ Project/ Code

Jailbreaking LLMs via complex ciphers overview

When "Competency" in Reasoning Opens the Door to Vulnerability: Jailbreaking LLMs via Novel Complex Ciphers

Divij Handa, Zehua Zhang, Amir Saeidi, Shrinidhi Kumbhar, Md Nayem Uddin, Aswin RRV, Chitta Baral

Reliable ML Workshop at NeurIPS 2025

Paper

UnSeenTimeQA overview

UnSeenTimeQA: Time-Sensitive Question-Answering Beyond LLMs' Memorization

Md Nayem Uddin, Amir Saeidi, Divij Handa, Agastya Seth, Tran Cao Son, Eduardo Blanco, Steven R. Corman, Chitta Baral

ACL 2025

Paper/ Code

MLLM evaluation overview

Evaluating Multimodal Large Language Models Across Distribution Shifts and Augmentations

Aayush Atul Verma*, Amir Saeidi*, Shamanthak Hegde*, Ajay Therala*, Fenil Denish Bardoliya*, Nagaraju Machavarapu*, Shri Ajay Kumar Ravindhiran*, Srija Malyala*, Agneet Chatterjee*, Yezhou Yang, Chitta Baral

EVGENFM Workshop at CVPR 2024

Paper

Hallucinations in negation tasks overview

Investigating and Addressing Hallucinations of LLMs in Tasks Involving Negation

Neeraj Varshney, Satyam Raj, Venkatesh Mishra, Agneet Chatterjee, Amir Saeidi, Ritika Sarkar, Chitta Baral

TrustNLP Workshop at NAACL 2025

Paper

DPO insights

Insights into Alignment: Evaluating DPO and its Variants Across Multiple Tasks

Amir Saeidi, Sivanshu Verma, Md Nayem Uddin, Chitta Baral

SRW Workshop at ACL 2025

Paper