Copertina

The Geometric Canary: Predicting Steerability and Detecting Drift via Representational Stability

Fuente: arXiv

Salvato in:

Dettagli Bibliografici
Autore principale:	Raju, Prashant C.
Natura:	Preprint
Pubblicazione:	2026
Soggetti:	Machine Learning Computation and Language
Accesso online:	Acceder al recurso
Tags:	Aggiungi Tag Nessun Tag, puoi essere il primo ad aggiungerne!!

Documenti analoghi

Geometric Stability: The Missing Axis of Representations
di: Raju, Prashant C.
Pubblicazione: (2026)

Exploring Chain-of-Thought Reasoning for Steerable Pluralistic Alignment
di: Zhang, Yunfan, et al.
Pubblicazione: (2025)

Bayesian Preference Learning for Test-Time Steerable Reward Models
di: Hong, Jiwoo, et al.
Pubblicazione: (2026)

A Course Correction in Steerability Evaluation: Revealing Miscalibration and Side Effects in LLMs
di: Chang, Trenton, et al.
Pubblicazione: (2025)

Integrating Time Series into LLMs via Multi-layer Steerable Embedding Fusion for Enhanced Forecasting
di: Chen, Zhuomin, et al.
Pubblicazione: (2025)

Inner Speech as Behavior Guides: Steerable Imitation of Diverse Behaviors for Human-AI coordination
di: Trivedi, Rakshit, et al.
Pubblicazione: (2026)

PRISM: A Geometric Risk Bound that Decomposes Drift into Scale, Shape, and Head
di: Lin, Chieh-Yen, et al.
Pubblicazione: (2026)

The Geometric Alignment Tax: Tokenization vs. Continuous Geometry in Scientific Foundation Models
di: Raju, Prashant C.
Pubblicazione: (2026)

Conditional Language Policy: A General Framework for Steerable Multi-Objective Finetuning
di: Wang, Kaiwen, et al.
Pubblicazione: (2024)

Event2Vec: A Geometric Approach to Learning Composable Representations of Event Sequences
di: Sulc, Antonin
Pubblicazione: (2025)

Tool Calling is Linearly Readable and Steerable in Language Models
di: Wu, Zekun, et al.
Pubblicazione: (2026)

Understanding Knowledge Drift in LLMs through Misinformation
di: Fastowski, Alina, et al.
Pubblicazione: (2024)

The Geometric Price of Discrete Logic: Context-driven Manifold Dynamics of Number Representations
di: Zhang, Long, et al.
Pubblicazione: (2026)

Breaking Quadratic Barriers: A Non-Attention LLM for Ultra-Long Context Horizons
di: Kiruluta, Andrew, et al.
Pubblicazione: (2025)

XLGoBench: Detecting cross-lingual skill gaps with algorithmic tasks
di: Jain, Purvam, et al.
Pubblicazione: (2026)

S2 Chunking: A Hybrid Framework for Document Segmentation Through Integrated Spatial and Semantic Analysis
di: Verma, Prashant
Pubblicazione: (2025)

Drifting Objectives for Refining Discrete Diffusion Language Models
di: Oba, Daisuke, et al.
Pubblicazione: (2026)

Inductive Entity Representations from Text via Link Prediction
di: Daza, Daniel, et al.
Pubblicazione: (2020)

The Canary's Echo: Auditing Privacy Risks of LLM-Generated Synthetic Text
di: Meeus, Matthieu, et al.
Pubblicazione: (2025)

The Confidence Manifold: Geometric Structure of Correctness Representations in Language Models
di: Cho, Seonglae, et al.
Pubblicazione: (2026)

Drift-to-Action Controllers: Budgeted Interventions with Online Risk Certificates
di: Lamaakal, Ismail, et al.
Pubblicazione: (2026)

InnerThoughts: Disentangling Representations and Predictions in Large Language Models
di: Chételat, Didier, et al.
Pubblicazione: (2025)

Joint Detection of Fraud and Concept Drift inOnline Conversations with LLM-Assisted Judgment
di: Senol, Ali, et al.
Pubblicazione: (2025)

Scale Determines Whether Language Models Organize Representation Geometry for Prediction
di: Xu, Weilun
Pubblicazione: (2026)

Foundation Model's Embedded Representations May Detect Distribution Shift
di: Vargas, Max, et al.
Pubblicazione: (2023)

Cross-Lingual Prompt Steerability: Towards Accurate and Robust LLM Behavior across Languages
di: Zhang, Lechen, et al.
Pubblicazione: (2025)

Protect$^*$: Steerable Retrosynthesis through Neuro-Symbolic State Encoding
di: Sathyanarayana, Shreyas Vinaya, et al.
Pubblicazione: (2026)

Stabilizing Policy Optimization via Logits Convexity
di: Chen, Hongzhan, et al.
Pubblicazione: (2026)

Silence the Judge: Reinforcement Learning with Self-Verifier via Latent Geometric Clustering
di: Zhang, Nonghai, et al.
Pubblicazione: (2026)

A Lightweight Method to Disrupt Memorized Sequences in LLM
di: Prashant, Parjanya Prajakta, et al.
Pubblicazione: (2025)

Patent Representation Learning via Self-supervision
di: Zuo, You, et al.
Pubblicazione: (2025)

Fair Text Classification via Transferable Representations
di: Leteno, Thibaud, et al.
Pubblicazione: (2025)

Objective Matters: Fine-Tuning Objectives Shape Safety, Robustness, and Persona Drift
di: Vennemeyer, Daniel, et al.
Pubblicazione: (2026)

Few-Shot Detection of Machine-Generated Text using Style Representations
di: Soto, Rafael Rivera, et al.
Pubblicazione: (2024)

Demystifying When Pruning Works via Representation Hierarchies
di: He, Shwai, et al.
Pubblicazione: (2026)

I Can't Believe It's Not Robust: Catastrophic Collapse of Safety Classifiers under Embedding Drift
di: Sahoo, Subramanyam, et al.
Pubblicazione: (2026)

Implicit Geometry of Next-token Prediction: From Language Sparsity Patterns to Model Representations
di: Zhao, Yize, et al.
Pubblicazione: (2024)

Mitigating Semantic Drift: Evaluating LLMs' Efficacy in Psychotherapy through MI Dialogue Summarization
di: Kumar, Vivek, et al.
Pubblicazione: (2025)

Model-Agnostic Sentiment Distribution Stability Analysis for Robust LLM-Generated Texts Detection
di: Li, Siyuan, et al.
Pubblicazione: (2025)

Methods for Generating Drift in Text Streams
di: Garcia, Cristiano Mesquita, et al.
Pubblicazione: (2024)