# Rotalabs - AI agent reliability research
# https://rotalabs.ai

## About
Rotalabs is the open-research division of Rota, Inc. We study how AI agents fail when they act in the world - and publish the benchmarks, probes, monitors, and protocols to measure it. Every result is published, installable, and reproducible from the raw trajectories up.

## Flagship study
- GlassBox: a pre-registered cross-pattern audit of linear probe gates for irreversible agentic actions. A null result, reported in full (benchmark, trajectories, trained probes, audit pipeline). Preprint coming soon; no arXiv ID yet. /research/glassbox/

## Published papers (arXiv)
- Field-Theoretic Memory for AI Agents - arXiv 2602.21220
- Quality-Diversity Evolution for Discovering Diverse Vulnerabilities in LLM Safety - arXiv 2606.00801 (ICLR 2026, Agents in the Wild workshop)
- Cross-Generational Transfer of Adversarial Attacks Reveals Non-Monotonic Safety Alignment in LLMs - arXiv 2606.00813

## Research areas (9)
- Agent Reliability: detection and verification for autonomous AI
- AI Evaluation Science: adversarial methods that resist gaming
- Memory Systems: field-theoretic memory (published - arXiv 2602.21220)
- Reasoning Verification: verifying outputs without ground truth
- Interpretability: practical methods for production AI
- Multi-Agent Trust: trust dynamics in multi-agent systems
- Adversarial Robustness for Agents: attack taxonomies and defenses
- Uncertainty Quantification: calibrated confidence for decision support
- World Models & Physical AI: embodied, physics-aware reasoning

## Open-source packages (12, AGPL-3.0, on PyPI and npm)
Core: rotalabs-context, rotalabs-probe, rotalabs-steer, rotalabs-verity,
rotalabs-cascade, rotalabs-redqueen, rotalabs-graph, rotalabs-fieldmem
Infrastructure: rotalabs-eval, rotalabs-accel, rotalabs-comply, rotalabs-audit

## Key pages
- Research: /research/
- GlassBox: /research/glassbox/
- Research Agenda: /research/agenda/
- Papers: /papers/
- Open Source: /open-source/
- Blog: /blog/
- About: /about/
- Contact: /contact/

## Contact
- General: hello@rotalabs.ai
- Research: research@rotalabs.ai
- Press: press@rotalabs.ai
- Address: 2261 Market Street STE 22728, San Francisco, CA 94114
- Phone: +1 (650) 899-9910

## Links
- Website: https://rotalabs.ai
- GitHub: https://github.com/rotalabs
- X: https://x.com/rotalabs
- PyPI: https://pypi.org/user/rotalabs/
- npm: https://www.npmjs.com/org/rotalabs

## Parent company
Rota, Inc. (Delaware C-Corporation)
- Rotalabs (rotalabs.ai) - open research on AI trust and reliability
- Rotascale (rotascale.com) - enterprise AI infrastructure
- Rotastellar (rotastellar.com) - aerospace and space AI systems