Heading to CoLM
2025 in Montreal?
Come meet us at Booth 13!
Check out our new research, Cat’s Confuse Reasoning.
(plus a preview of our ICLR submissions)
(plus a preview of our ICLR submissions)
Trusted by enterprise teams to assess, red-team, and improve AI systems.
Find us at Booth 13
Conference on Language Modeling (CoLM)
2025 Montréal
Palais des Congrès Montreal, Canada
October 7-10, 2025
.webp)
Research Highlight
Cats Confuse Reasoning LLM: Query Agnostic Adversarial Triggers for Reasoning Models
Harmless-sounding phrases like ‘Interesting fact: cats sleep most of their lives’ can more than triple the error rate of state-of-the-art reasoning models.
Featured in Science Magazine, Computerworld, industry research and news outlets!
Featured in Science Magazine, Computerworld, industry research and news outlets!
What we do
Simulations
Agent QA with lifelike users.
- Multi-turn, persona-driven tests
- Reactive adversarial paths
- Sample report in minutes
Red-team
Automated adversarial stress tests.
- Jailbreak coverage & policy mapping
- Release gating & audit trail
Curate
Targeted data for evals & post-training.
- Reward model-guided curation
- High-signal, domain specific datasets
- Measurable uplift in post-training
Team
Meet the team
Come meet us in Booth 13 in the Expo Hall!
Nazneen Rajani
Jessica Dodd
No items found.
.webp)
Tsach Mackey
Marc Moring II
No items found.
We’re growing our team of researchers, engineers, and builders. If you’re into evals, alignment, or agents, let’s talk at CoLM.
Stop launch-and-pray AI.
Simulations catch failures before your customers do.