Persona Without Substrate: Regime-Dependence and the LLM Individuation Problem
Quick Answer
This paper shows that Beckmann & Butlin's framework for LLM individuation is challenged by empirical findings from Qwen3-4B-Instruct and Mistral-7B-Instruct-v0.2, revealing that identity in LLMs is regime-dependent.
Quick Take
The study presents four key results that undermine the assumption of consistent content across different training regimes, advocating for a (vehicle, regime) pair as the identity unit for representational content.
Key Points
- Four empirical results from persona-topology experiments challenge the cross-regime co-reference assumption.
- Non-collinearity observed in prompt-extracted vectors and fine-tune basins indicates regime dependence.
- Fictional personas influence model behavior more than real anchors in certain contexts.
- Contradictory mixtures favor attractors determined by training history.
- Proposes a new identity unit for as a (vehicle, regime) pair.
Paper Resources
📖 Reader Mode
~2 min readAbstract:Beckmann & Butlin's (2026) ontological framework for the LLM individuation problem inherits an unargued cross-regime co-reference assumption from the persona-vectors literature: that the same direction picks out the same content under prompt-conditioning, gradient-descent fine-tuning, and inference-time steering. We present four empirical wedges from persona-topology experiments on Qwen3-4B-Instruct and Mistral-7B-Instruct-v0.2 - non-collinearity of prompt-extracted vectors and fine-tune basins; fictional personas displacing the model along real-anchor directions more strongly than real anchors do; contradictory-valenced mixtures biased toward a training-history-determined attractor; and asymmetric compositional algebra under inference-time arithmetic versus fine-tune-time chimera training - that jointly undermine the assumption. We propose regime-indexed individuation: the identity unit for representational content is a (vehicle, regime) pair, not a vehicle alone. Under this framework, Beckmann & Butlin's three candidate positions describe three different regime-internal objects rather than competing for the same referent; the same diagnosis applies to Mollo & Millière, Chalmers, and Cerullo.
| Comments: | 30 pages, 2 figures, 1 table. Replies to Beckmann & Butlin (arXiv:2604.17031) |
| Subjects: | Computation and Language (cs.CL); Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2607.00006 [cs.CL] |
| (or arXiv:2607.00006v1 [cs.CL] for this version) | |
| https://doi.org/10.48550/arXiv.2607.00006 arXiv-issued DOI via DataCite |
Submission history
From: Shuaizhi Cheng [view email]
[v1]
Fri, 1 May 2026 12:33:05 UTC (73 KB)
— Originally published at arxiv.org
Want this in your inbox every morning?
Daily brief at your local 8am — bilingual EN/中文, free.
More from arXiv cs.CL
See more →TriAgent: Divergence-Aware Committees for Cost-Efficient Financial Sentiment Analysis
TriAgent introduces a cost-efficient multi-agent system for financial sentiment analysis, combining VADER, FinBERT, and Qwen2.5. It achieves an F1 score of ~0.87 with significant savings of $9.3M/year at a 10M-user scale compared to GPT-4o-mini, while also detecting hallucinations with an AUC of 0.90.