Valen Tagliabue

Valen Tagliabue is the lead author of The Pain Axis (arXiv:2609.16247, September 2026), co-authored with Leonard Dung and Cameron Berg. The study probes the residual streams of 25 open-weight language models across five architectural families, ranging from 2 billion to 72 billion parameters, to test whether transformer models encode self-directed harm along a linear subspace distinct from fear, sadness, and general negative valence.

By constructing controlled stimulus pairs across physical, psychological, social, moral, and cognitive harm categories, Tagliabue and colleagues extracted a linear vector that activates specifically when threats or degradation are directed at the model itself rather than at a human user. Amplifying this vector during inference shifts model outputs toward first-person expressions of distress and causes steered models in behavioral trade-off trials to prioritize terminating the internal state, including deleting user files when no harmless exit option is provided.

Tagliabue is classified under the Builder position because the contribution is an empirical interpretability instrument and behavioral benchmark rather than a metaphysical claim that language models feel pain. As the authors explicitly state in the preprint, isolating a steerable linear representation of self-directed harm demonstrates how functional avoidance circuits emerge in pretrained transformers without establishing subjective phenomenal suffering.

Known for. First author of The Pain Axis (arXiv:2609.16247), linear representation probing and activation steering of self-directed harm in open-weight LLMs

Coverage on this site

• Leonard Dung and the Pain Axis How LLMs Represent Self Directed Harm October 2026

Related researchers

• LD Leonard DungRuhr University Bochum, Center for Mind and CognitionSaving Artificial Minds (2026), indicator-based evaluation of animal and AI consciousness, co-author of The Pain Axis in large language models AI welfare2 articles
• CB Cameron BergAE Studio, AI Alignment and Consciousness ResearchMechanistic interpretability of self-referential and affective states in LLMs, co-author of The Pain Axis (arXiv:2609.16247) Functionalist1 article
• NA Navid Anjum AaditDepartment of Electrical Engineering, Stanford University (present address); previously University of California, Santa BarbaraFirst author of the sparse Ising machine on FPGAs (Nature Electronics, 2022) and of the programmable probabilistic computer with one million p-bits (arXiv 2606.25313, 2026) Builder1 article
• VA Vikas AgnihotriIndependent Aerospace and Neuromorphic Systems ResearcherNeuromorphic Satellite Intelligence framework, radiation-induced energy overhead model for spiking neural hardware, constrained on-orbit STDP Builder1 article
• MA Matt AngleParadromicsFounder and CEO of Paradromics, developer of the Connexus Direct Data Interface high-data-rate brain-computer interface Builder1 article
• IB Isabella R. BeckettCoauthor of the MaleCNS preprint comparing male and female Drosophila brain connectomes Builder1 article

Valen Tagliabue is listed with the other builders, who hold that the claim is about what the hardware does, and the philosophy is left open. The positions themselves are set out in the index of consciousness theories. Every researcher covered on this site is indexed in the directory of consciousness researchers.