To learn more:
Website: agentcoma.github.io
Preprint: arxiv.org/abs/2508.19988
A big thanks to my brilliant coauthors Lihu Chen, Ana Brassard, @joestacey.bsky.social, @rahmanidashti.bsky.social and @marekrei.bsky.social!
Note: We welcome submissions to the #AgentCoMa leaderboard from researchers š
At #NeurIPS2025 today, @lisaalaz.bsky.social is presenting our joint paper on Reverse Engineering Human Preferences with Reinforcement Learning! Demonstrating undetectable attacks on LLM-as-a-judge benchmarks. Great collaboration with
@cohereforai.bsky.social and a well-deserved NeurIPS spotlight!
We test AgentCoMa on 61 contemporary LLMs of different sizes, including reasoning models (both SFT and RL-tuned). While the LLMs perform well on commonsense and math reasoning in isolation, they are far less effective at solving AgentCoMa tasks that require their composition!
We have released #AgentCoMa, an agentic reasoning benchmark where each task requires a mix of commonsense and math to be solved š§
LLM agents performing real-world tasks should be able to combine these different types of reasoning, but are they fit for the job? š¤
š§µā¬ļø
AgentCoMa is an Agentic Commonsense and Math benchmark where each compositional task requires both commonsense and mathematical reasoning to be solved. The tasks are set in real-world scenarios:ā¦
agentcoma.github.io
In contrast, we find that:
- LLMs perform relatively well on compositional tasks of similar difficulty when all steps require the same type of reasoning.
- Non-expert humans with no calculator or internet can solve the tasks in #AgentCoMa as accurately as the individual steps.
We also observe that LLMs fail to activate all the relevant neurons when they attempt to solve the tasks in Agent-CoMa. Instead, they mostly activate neurons relevant to only one reasoning type, likely as a result of single-type reasoning patterns reinforced during training.
So why do LLMs perform poorly on the apparently simple tasks in #AgentCoMa?
We find that tasks combining different reasoning types are a relatively unseen pattern for LLMs, leading the models to contextual hallucinations when presented with mixed-type compositional reasoning.