|
Akshat Gupta
I am a third-year PhD student at UC Berkeley affiliated with BAIR and Berkeley Speech Group, advised by Gopala Anumanchipalli.
Before joining Berkeley, I spent two wonderful years at AI Research, JPMorgan where I worked as an NLP Research Engineer.
Currently I am actively working on the following two problems:
Some of my past research interests that I still think about from time to time:
- Lifelong Knowledge editing [1, 2, 3, 4, 5]
- Interpretability [1, 2, 3]
- LLMs and Poker [1, 2]
- Generative Spoken Language Modelling [1, 2, 3, 4]
- Personality Measurement in LLMs [1, 2, 3]
Feel free to reach out to me if you want to chat about any of my current or past interests. I also love a good conversation about AI, technology and where all of this is headed.
Email  / 
Google Scholar  / 
Twitter  / 
Linkedin  / 
CV (Updated Feb 2026)
|
|
Blog
I recently wrote a blog where I reviewed one of the AI generated papers by Sakana AI. If you're interest, here's the link.
News
- March 3, 2026 - SkyDiscover is released.
- Feb 23, 2026 - "AdaEvolve: Adaptive LLM Driven Zeroth-Order Optimization" is released on Arxiv.
- Jan 28, 2026 - "Evolutionary Strategies lead to Catastrophic Forgetting in LLMs" is released on Arxiv.
- Jan 4, 2026 - "Geometric Interpretation of Layer Normalization and a Comparative Analysis with RMSNorm" accepted to EACL 2026 Findings
- Oct 22, 2025 - "How LLMs use their Depth?" is released on Arxiv. This AI-generated video does a great job summarizing our work.
- Aug 20, 2025 - "Lifelong Knowledge Editing requires Better Regularization" accepted to EMNLP 2025 Findings
- May 19, 2025 - Started internship at IBM Research.
- May 15, 2025 - "Efficient Knowledge Editing via Minimal Precomputation" accepted to ACL 2025 Main Conference
- March 4, 2025 - Outstanding Paper Award at the KnowFM Workshop @ AAAI 2025. (Paper Link)
- Feb 2025 - I will be attending AAAI 2025 in Philadephia from Feb 26 - March 4. Please reach out if you want to chat about interpretability, reasoning or knowlegde editing.
- Jan 22, 2025 - "Sylber: Syllabic Embedding Representation of Speech from Raw Audio" accepted to ICLR 2025
- Dec 9, 2024 - "PokerBench: Training LLMs to become Professional Poker Players" accepted to AAAI 2025
- Nov 2024 - I will be attending EMNLP 2024 in Miami from November 11 - 16. Please reach out if you want to chat.
- Sep 20, 2024 - "Rebuilding ROME : Resolving Model Collapse during Sequential Editing" accepted to EMNLP 2024 Main Conference
- Sep 20, 2024 - "A Unified Framework for Model Editing" accepted to EMNLP 2024 Findings
- Sep 20, 2024 - "Self-Assessment Tests are Unreliable Measures of LLM Personality" accepted to BlackboxNLP 2024, co-located with EMNLP 2024
- August 12, 2024 - Attending ACL 2024 in Bangkok!!
- May 15, 2024 - "Model Editing at Scale leads to Gradual and Catastrophic Forgetting" accepted to ACL 2024 Findings
|