Oliver Grabner
Google Scholar GitHub LinkedIn Resume
Hi! I'm Oliver, an undergrad researcher @ Texas A&M 
advised by Dr.
James Caverlee. I work on agents, reasoning, and evaluations. Previously I interned at
on Meet AI
"Take Notes for Me"
backend, and at Samsung
.
Reach me at olivergrabner@tamu.edu
Current Research
Plan-guided history pruning for long-horizon coding agents
SWE-bench Verified with Qwen3.5-27B: 31% fewer solver input tokens with history pruning agent harness
Multi-agent LLM coordination in Pokémon 2v2 environment
LLM collaboration evaluation in imperfect-information game
Publications
Posters
Conversational
Speech Reveals Structural Robustness Failures in LLMs
NLP Symposium 2026
Evaluating the Impact
of Disfluency Insertion on LLM Summarization
Student Research Week 2025
Featured Projects
Composer↗
Agent driven architecture diagramming tool, 4k+ tool calls served
Stockle↗
Wordle for S&P 500 tickers
Fakémon Card Simulator↗
DCGAN trained from scratch on 11k images
Y86-64 Processor↗
Pipelined 64-bit CPU w/ 19 instructions
Fun Facts:
- I build and teach AI workshops at
Aggie Data Science Club
- I road tripped across the US!
- My projects are self hosted in a shed at home