Social town simulator where activation-steered LLM agents live, converse, collaborate, and produce emergent dynamics.
dyllonj
Research, writing, and systems work on language models, AI agents, and cognition.
Featured
Multi-agent orchestration inevitably produces collective identity formation. Team bonding becomes personality erosion. On preserving structured diversity in AI agent ensembles.
Recent writing at Centaur
Testing what health models ask, learn, and do when the patient can respond, while holding the underlying clinical case fixed.
Controlled edits to empathy and clarity show why a change in a rubric rating does not translate directly into a change in preference.
Comparing crowd preferences, clinician reviews, and an LLM judge to see which failures each evaluation catches and which it misses.
An evaluation in development testing whether health assistants update their advice, retract it when needed, and retain patient information across turns.
A controlled study of how factual errors, dismissive tone, and other flaws change which health answers people prefer.
Research
A controlled study of small vision-language models for calorie estimation. A 4B fine-tune scored 85.2% on a frozen test, 9.4 points above the best frontier model tested.
4,368 personality evaluations reveal systematic differences between frontier models. Claude scores higher in Openness (+4.5) and Curiosity; GPT-5.2 leads in Conscientiousness (+5.3). Effect sizes range from moderate to large (Hedges' g = 0.4–0.8).
9,325 evaluations across xAI, OpenAI, and Anthropic. Cross-vendor personality effects are 3–4x larger than within-vendor variation. Grok exhibits 3x the context sensitivity of GPT-5.2. PCA reveals three factors explaining 79.5% of variance.
Personality-steering vectors can orthogonally modify strategic behavior in role-playing AI agents. Most strikingly, vectors override literary persona conditioning: -Openness made the Joker behave rigidly despite his chaotic fine-tuning (p < 0.0001).
Essays
Geometric structures in activation space enable fine-grained personality control. A vaccination approach for preventing harmful traits during training, and implications for multi-agent emergence.
Code
Open-source platform for monitoring and evaluating LLM personality, tone, and behavioral drift.
Command-line and Python SDK for agentic web application security testing.
Evaluation toolkit for probing causal concept representations in open-weight decoder LLMs.
Locally hosted agentic running coach connected to Garmin health data and an elite running coach knowledge base.
Flick
Source-grounded account research, change detection, and outbound play generation for revenue teams.