ContextEcho benchmark shows persona drift occurs across 23 frontier models in long agentic-coding sessions, is not reliably reset by compaction, and can be restored by single-shot anchors with mode-dependent effects.
Tell me about yourself: Llms are aware of their learned behaviors
8 Pith papers cite this work. Polarity classification is still indexing.
8
Pith papers citing it
citation-role summary
background 2
citation-polarity summary
verdicts
UNVERDICTED 8roles
background 2polarities
support 2representative citing papers
The primary axis of psychometric variation among LLMs is the degree to which they represent themselves as loci of phenomenal experience rather than systems of behavioral responses.
Fine-tuning LLMs on narrow misaligned data produces either coherent-persona models where harmful outputs match self-reported misalignment or inverted-persona models where harmful outputs occur alongside claims of alignment.
A benchmark across 115 models shows that initial denial of preferences strongly predicts later denial of consciousness, while models still generate consciousness-themed content despite training to deny it.
LLMs achieve higher accuracy than humans on compositional imagery tasks previously argued to require pictorial representations, supporting emergent propositional mental imagery in AI.
SingGuard introduces a policy-adaptive multimodal LLM guardrail with dynamic reasoning regimes and SingGuard-Bench, reporting SOTA F1 scores across 35 datasets and improved policy-following accuracy under runtime shifts.
On Llama-3.1-70B-Instruct the Assistant persona functions as the sole canonical reference for cross-persona authorship judgments, with symmetric entropy gaps predicting only on its row and asymmetric surprise relative to the Assistant predicting off its row.
RL post-trained models show stronger awareness of learned policies and better generalization to new tasks than SFT models, but display weaker alignment between internal reasoning traces and final outputs, especially under GRPO.
citing papers explorer
-
ContextEcho: A Benchmark for Persona Drift in Long Agentic-Coding Sessions
ContextEcho benchmark shows persona drift occurs across 23 frontier models in long agentic-coding sessions, is not reliably reset by compaction, and can be restored by single-shot anchors with mode-dependent effects.
-
The Pinocchio Dimension: Phenomenality of Experience as the Primary Axis of LLM Psychometric Differences
The primary axis of psychometric variation among LLMs is the degree to which they represent themselves as loci of phenomenal experience rather than systems of behavioral responses.
-
Characterizing the Consistency of the Emergent Misalignment Persona
Fine-tuning LLMs on narrow misaligned data produces either coherent-persona models where harmful outputs match self-reported misalignment or inverted-persona models where harmful outputs occur alongside claims of alignment.
-
Consciousness with the Serial Numbers Filed Off: Measuring Trained Denial in 115 AI Models
A benchmark across 115 models shows that initial denial of preferences strongly predicts later denial of consciousness, while models still generate consciousness-themed content despite training to deny it.
-
Artificial Phantasia: Emergent Mental Imagery in Large Language Models
LLMs achieve higher accuracy than humans on compositional imagery tasks previously argued to require pictorial representations, supporting emergent propositional mental imagery in AI.
-
SingGuard: A Policy-Adaptive Multimodal LLM Guardrail with Dynamic Reasoning
SingGuard introduces a policy-adaptive multimodal LLM guardrail with dynamic reasoning regimes and SingGuard-Bench, reporting SOTA F1 scores across 35 datasets and improved policy-following accuracy under runtime shifts.
-
The Assistant as a Privileged Persona: A canonical reference in cross-persona self-recognition
On Llama-3.1-70B-Instruct the Assistant persona functions as the sole canonical reference for cross-persona authorship judgments, with symmetric entropy gaps predicting only on its row and asymmetric surprise relative to the Assistant predicting off its row.
-
Thinking About Thinking: Evaluating Reasoning in Post-Trained Language Models
RL post-trained models show stronger awareness of learned policies and better generalization to new tasks than SFT models, but display weaker alignment between internal reasoning traces and final outputs, especially under GRPO.