Abaka AI @ COLM 2026 — Booth #106
Find us at COLM 2026, October 6–8, at the Hilton San Francisco Union Square — Booth #106, Plaza Ballroom.
Post-training data, RL environments, expert annotation: built to your spec by the data partner behind Terminal-Bench, SuperGPQA & FormalMATH.
What we work on:
Post-Training Data — SFT and RLHF datasets for reasoning, coding, and agentic tasks.
RL Environments — MCP-based, stateful and reproducible, with verifiable and rubric-graded rewards.
Expert Annotation — domain experts for preference ranking and rubric scoring.
Model Evaluation — custom benchmarks, red-teaming, and evals.
If your next training run needs data your pipeline can't produce, come find us.
And save Wednesday night, Oct 7 for La Dolce Vita with Abaka AI at a secret location, a 1-minute walk from the conference. Request your invite here.
About Abaka AI: Abaka AI is a full-lifecycle data partner for AI teams — data collection, curation, annotation, and model evaluation across text, image, audio, video, 3D, and agent/RL data. We work with 1,000+ AI labs, partners, and research groups across autonomous driving, generative AI, and embodied AI. Follow us on LinkedIn and X.