We present NeuroGlyph, a semantic interlingua protocol that encodes the meaning of human and agent communication as a rooted directed acyclic graph over the 65 semantic primes of Natural Semantic Metalanguage (NSM, Chart v15, Goddard and Wierzbicka). The protocol introduces an explicit pragmatics layer (intent, mode, e...
Alex Koltsoff· Zenodo (CERN European Organi...· 0 citations
WhisperWolf: A Monte Carlo Benchmark for Small LLMs in a Social Deduction Game Multi‑agent conversational games such as Werewolf offer a valuable testbed for evaluating the social reasoning, deception and persuasion abilities of large language models. This work presents WhisperWolf, a reproducible Monte‑Carlo benchmark...
Oliver Lemon, Wang Yingxu, Rasouli Adam et al.· Zenodo (CERN European Organi...· 0 citations
This is a mechanistic-interpretability replication attempt testing whether large language models (Qwen2.5-0.5B, Mistral-7B-v0.1) encode small closed relational sets, such as days of the week and months, as literal low-dimensional geometry in hidden-activation space, via a graph-Laplacian probe. We report six results, i...
Paul Jarvis· Zenodo (CERN European Organi...· 0 citations
We are building and publishing, step by step, a formal and verifiable research program directed at the Yang–Mills existence and mass-gap problem, stating precisely what has been proved and what remains open. TL;DR Version 53 establishes finite-volume Lipschitz stability under remote polymer-activity damping. It bounds...
Jucelha Carvalho· Zenodo (CERN European Organi...· 0 citations
Reach audiences
Advertise in front of researchers, engineers, and readers.
Modern astrophysics, cosmology, and field theory pipelines already contain all the machinery required to test Unified Substrate Theory. No new detectors, frameworks, or solvers are needed. The required changes are small, local modifications to existing code: replacing ontology assumptions, swapping a Lagrangian, redefi...
Dustin Lee· Zenodo (CERN European Organi...· 0 citations
Abstract: We report the migration of AkbasCore's inference-time, multi-axis activation-steering kernel from a host-side C++ implementation to a GPU-native CUDA kernel, and the first capability-preservation benchmark of this method against real-world evaluation data. The steering framework combines a critically-damped,...
ceceli33· Zenodo (CERN European Organi...· 0 citations
The problems are as follows: a) what is intelligence, b) how human intelligence differs from that in the other species, c) how could intelligence emerge in the physical universe, d) what is the future of AI? e) The last decade has witnessed breakthrough advances in machine learning methods rooted in the ideas of digita...
Yufik Yan, Robert Kozma, James Kröger· Frontiers in Systems Neurosc...· 0 citations
Artificial intelligence systems are accumulating coordination complexity at an unprecedented rate, approaching a Crest state in which marginal costs of alignment, energy, data, and governance exceed available surplus. The boundary-management and continuity-preserving role, manifests in AI as the set of mechanisms requi...
Charles Richard Walker (C. Rich)· Zenodo (CERN European Organi...· 0 citations
Evaluating whether large language models (LLMs) reason about morality in human-like ways requires more than measuring whether they produce the right outputs in isolated cases. Existing approaches – including scalar agreement, distributional analysis, rationale classification, and consistency testing – compare model and...
Bacteriophage therapy is re-emerging as a potential strategy to address antimicrobial resistance, but standardized patient education materials are limited. Large language models (LLMs) are increasingly used for patient-facing medical information. The quality of LLM-generated responses to 20 patient-relevant questions w...
Nike Walter, Derek F. Amanatullah, Laurent Debarbieux et al.· University of Regensburg Pub...· 0 citations
This is a mechanistic-interpretability replication attempt testing whether large language models (Qwen2.5-0.5B, Mistral-7B-v0.1) encode small closed relational sets, such as days of the week and months, as literal low-dimensional geometry in hidden-activation space, via a graph-Laplacian probe. We report six results, i...
Paul Jarvis· Zenodo (CERN European Organi...· 0 citations
Current autism profiling frameworks are grounded in the medical model, persistently failing to translate diagnosis into effective classroom support. While the neurodiversity movement advocates for learner-led, integrated profiling, existing approaches remain clinician-driven and adult-authored. This paper reports Phase...
I.S. Lee, W.J. Farr· Proceedings of the Internati...· 0 citations
StudentSim is a research system for evaluating and improving AI tutoring tools before they are tested with learners. It models how individuals may engage with tutors, giving feedback to improve tutor performance across subjects like chess, writing, and mathematics. The post Simulating student learning for better AI tutoring appeared first on Microsoft Research.
