Helena Salgueiro

Research

Helena leads cantor-18, the cell working on specification refinement from traces. The cell's premise is that formal specifications inferred from observed agent traces can be used to gate future runs, replacing the need for a hand-written specification at the moment a new behaviour appears.

She has been with alphabell since 2021 and is a co-author of the verifiable-policies result. She is one of the contributors who works comfortably at the boundary between formal methods and machine learning; the cantor-18 line of work uses tooling from both communities.

Helena is based in Lisbon. She runs an annual workshop on specifications-from-traces at CAV and is one of the lab's bridges to the formal-methods community.

Background

Ph.D. computer science, University of Lisbon, 2015.

Prior to alphabell: University of Lisbon; Cantor Initiative; Wayfarer Institute.

Selected publications

• Mar 2025 · ab-verifiable-pol
  Toward Formal Verification of Learned Policies in Bounded Environments
  Aviva Stern, Sun Kyung-min, Felipe Avelar
• Sep 2024 · ab-interpretabili
  Interpretability Cell Pairing: how every dual-use capability run gets a watchful sibling
  Karima Belkadi, Hester Vandekerckhove, Yuki Cho
• May 2025 · ab-mechanistic-ci
  Mechanistic Circuit Analysis at Frontier Scale: cells as a unit of interpretability
  Jiang Yifei, Nico Almgren, Karima Belkadi, Hester Vandekerckhove
• Sep 2025 · ab-scalable-overs
  Scalable Oversight for Multi-Step Agent Systems: a Debate-Plus-Trace Approach
  Ifeoma Nwosu-Howard, Hiroshi Tanigawa, Maral Lotfi, Ruth Wernicke
• Nov 2024 · ab-sandboxed-self
  Sandboxed Self-Modification: a confinement specification and implementation
  Liora Sabatini, Cheung Wai-Lin, Marek Holub

Full publications index →

Recent talks

• Specifications inferred from traces, FM 2025
• What a specification has to capture, ML Safety Workshop 2024