First systematic test shows activation steering robustness drops sharply (up to 64%) under adversarial input perturbations across multiple extraction methods, models, and personas.
Title resolution pending
8 Pith papers cite this work. Polarity classification is still indexing.
8
Pith papers citing it
citation-role summary
method 2
citation-polarity summary
years
2026 8roles
method 2polarities
use method 2representative citing papers
SMX explains spectral ML classifiers by summarizing expert zones with PCA, testing quantile predicates via perturbation, aggregating via directed graph centrality, and reconstructing thresholds back onto original spectra.
Data-similarity and data-influence produce significantly overlapping rankings of training documents for LLM outputs, with asymmetry allowing a favorable cost-accuracy trade-off.
Arbor combines a coordinator, executors, and a hypothesis tree to enable cumulative autonomous research, outperforming Codex and Claude Code by over 2.5x on six real tasks and reaching 86.36% Any Medal on MLE-Bench Lite.
Synthetically formalizing information needs into topics with descriptions and narratives improves LLM relevance assessor agreement with humans and reduces over-labeling of relevant documents on TREC Deep Learning and Robust04.
Relative Probability Association Metric (RPAM) measures LM associations via softmax-normalized continuation probabilities and correlates strongly with human associations and downstream LM behavior across three models.
Young adults engage with low-quality news content on social media despite stating preferences for high-quality, accurate, and diverse information, and they produce higher-quality feeds when curating for a hypothetical persona.
Visibility in generative AI search must be assessed as a distribution over repeated measurements rather than single queries because outputs vary across runs, prompts, and time.
citing papers explorer
-
Adversarial Robustness of Activation Steering in Large Language Models
First systematic test shows activation steering robustness drops sharply (up to 64%) under adversarial input perturbations across multiple extraction methods, models, and personas.
-
Spectral Model eXplainer: a chemically-grounded explainability framework for spectral-based machine learning models
SMX explains spectral ML classifiers by summarizing expert zones with PCA, testing quantile predicates via perturbation, aggregating via directed graph centrality, and reconstructing thresholds back onto original spectra.
-
Quantifying the Agreement Between Data-Influence and Data-Similarity to Understand LLM Behavior
Data-similarity and data-influence produce significantly overlapping rankings of training documents for LLM outputs, with asymmetry allowing a favorable cost-accuracy trade-off.
-
Toward Generalist Autonomous Research via Hypothesis-Tree Refinement
Arbor combines a coordinator, executors, and a hypothesis tree to enable cumulative autonomous research, outperforming Codex and Claude Code by over 2.5x on six real tasks and reaching 86.36% Any Medal on MLE-Bench Lite.
-
Formalized Information Needs Improve Large-Language-Model Relevance Judgments
Synthetically formalizing information needs into topics with descriptions and narratives improves LLM relevance assessor agreement with humans and reduces over-labeling of relevant documents on TREC Deep Learning and Robust04.
-
RPAM: A Principled Metric for Evaluating Associations in Language Models with High Predictive Validity in Downstream Outputs
Relative Probability Association Metric (RPAM) measures LM associations via softmax-normalized continuation probabilities and correlates strongly with human associations and downstream LM behavior across three models.
-
Understanding the Gap Between Stated and Revealed Preferences in News Curation: A Study of Young Adult Social Media Users
Young adults engage with low-quality news content on social media despite stating preferences for high-quality, accurate, and diverse information, and they produce higher-quality feeds when curating for a hypothetical persona.
-
Don't Measure Once: Measuring Visibility in AI Search (GEO)
Visibility in generative AI search must be assessed as a distribution over repeated measurements rather than single queries because outputs vary across runs, prompts, and time.