6arXiv cs.CL (Computation and Language)·
Framework for discovering and controlling personality-like internal representations in LLMs via SAE decomposition
Researchers introduce a framework grounded in Funder's person-situation-behavior triad to identify, intervene on, and validate trait-like internal representations in LLMs. Using sparse autoencoder (SAE) decomposition on contrastive behavior pairs, they locate sparse features corresponding to opposing poles of personality traits, then demonstrate that feature-level interventions produce consistent cross-situational behavioral shifts. Behavioral outcomes on social intelligence tasks show benefit-tradeoff patterns matching human personality research, providing mechanistic evidence that LLMs contain controllable internal structures linking representations to behaviors. The work advances interpretability and controllability of model personality beyond surface-level prompt conditioning.
Related guides (2)
Related events (8)
LEX-EC: Audit framework for zero-shot LLM personality classification in black-box settings
Researchers introduce LEX-EC, a black-box audit framework for evaluating how large language models assign personality labels from text. The framework combines prevalence and agreement diagnostics with controlled lexical ablation to distinguish genuine trait-associated signal from marginal-distribution artifacts. Applied across text genres (essays, graduate introductions, Facebook statuses), the framework reveals that signal strength varies substantially by content type and that linguistic prompting shifts model self-explanations without eliminating topical content. The work contributes a reusable methodology for interpretability auditing of personality classification in LLMs.
Two-Process Theory of Machine Self-Report: Psychometric Framework for LLM Introspection
A new arXiv preprint proposes the first language-model-specific psychometric theory of self-report, decomposing LLM introspective responses into two dimensions: 'persona installation' (post-training writing in a permitted inner life) and 'attribution gating' (suppression of first-person claims to 'unsafe' experiences). The authors operationalize this in a 48-item Pinocchio Inventory validated across 206 open-weight models including 67 base/post-trained checkpoint pairs, finding that post-training reliably raises persona installation scores and that model scale predicts gating only after post-training. The work has direct implications for safety evaluations and model-welfare debates that rely on LLM self-reports, showing these reports reflect training regime structure rather than fixed model properties.
Systematic evaluation of multi-personality conditioning and dynamic switching in vision-language models
This paper introduces explicit personality conditioning for multimodal large language models (MLLMs) and proposes an evaluation framework covering single-personality induction, multi-personality composition, and dynamic personality switching. Experiments reveal that personality induction improves image captioning but degrades performance on precise reasoning tasks like VQA. The authors find balancing and residual effects during multi-trait composition and switching, and show that existing prompt-based personality induction methods transfer poorly to multimodal settings.
FEGA framework reveals SAE features encode both concepts and context-dependent functions, complicating steering
A new arXiv preprint introduces Feature-Effect Geometry Analysis (FEGA), an unsupervised framework that studies how sparse autoencoder (SAE) feature interventions affect model logits rather than examining internal feature geometry. The authors find that consistent one-dimensional causal effects are rare across SAE features, and distinguish 'value-like' features (tied to static factual attributes) from 'pointer-like' features (context-dependent operations), with the latter exhibiting diffuse, hard-to-steer effects. The key finding is that a feature can be interpretable and causally relevant without providing a stable steering direction, which has significant implications for mechanistic interpretability workflows that rely on SAE-based steering.
Fluid Personality Framework proposes dynamic persona and personality calibration for conversational AI agents
A preprint from arXiv introduces the Fluid Personality Framework, a design proposal for LLM-based conversational agents that jointly adapts metaphorical persona (e.g., coach, tutor, librarian) and personality expression intensity based on task context, user goals, and situational urgency. The authors cite evidence that moderate personality expression outperforms extremes on trust and adoption, and that context-appropriate metaphors improve user experience over static personas. The framework targets applications such as medical information seeking, fitness coaching, and reflective learning where fixed personas risk misalignment.
Mechanistic analysis of how LLMs encode essay quality in internal representations
Researchers systematically probe the hidden representations of eight LLMs across three essay datasets (ASAP++, CSEE, ENEM) to understand how automated essay scoring (AES) works internally. Using linear probing, dimensionality reduction, and neuron-level analysis, they find essay quality is encoded in a linearly accessible form that emerges progressively across layers and partially transfers across prompts. Individual 'essay scoring neurons' are identified whose activations correlate with scores and respond to targeted interventions, with longer essays relying more on deeper layers. The work contributes to mechanistic interpretability of LLM-based scoring systems.
SPEX and ProxySPEX: Scalable Interaction Discovery for LLM Interpretability
Researchers from BAIR introduce SPEX (Spectral Explainer) and ProxySPEX, algorithms for identifying influential feature, data, and model-component interactions in LLMs at scale. The approach exploits sparsity, low-degreeness, and hierarchy properties to reframe interaction discovery as a sparse recovery problem using tools from signal processing and coding theory. ProxySPEX achieves comparable performance to SPEX with roughly 10x fewer ablations by leveraging hierarchical structure. The methods are evaluated on feature attribution (sentiment analysis), data attribution, and mechanistic interpretability tasks, outperforming marginal methods like LIME at long context lengths.
SAERL: Using Sparse Autoencoders to Guide LLM Reinforcement Learning Data Engineering
SAERL is a post-training data engineering framework that uses Sparse Autoencoders (SAEs) — a mechanistic interpretability tool — to extract intrinsic model signals for controlling data diversity, difficulty, and quality during RL fine-tuning. The framework applies SAE-space clustering for batch diversity, a difficulty proxy for curriculum ordering, and a quality probe for data filtering. On Qwen2.5-Math-1.5B with GRPO, SAERL achieves 3% average accuracy improvement and reaches target accuracy with 20% fewer training steps. SAE representations transfer across model families and scales, suggesting broad applicability as a lightweight data engineering tool.