Elicitation

Follow this topic Ask about this topic

Momentum

13 papers in the last four weeks, up 117% on the four weeks before. 0.1% of all new papers.

Jul 13Week of Sep 28

Latest papers 78

All topics

Learning Steganography Is Easy, Learning Steganographic Reasoning Is Hard
Sep 30, 2026Julian Schulz, Lukas Fülle, Rieke FruengelSteganography Chain-of-Thought Reasoning
Meridian Cambridge · SAIGE
Reach Into The CHOIR: Free-List Elicitation Uncovers Distinct Model Voices in LLM Ensembles
Sep 29, 2026Ben Wigler, Maria TsfasmanElicitation Prompt Engineering
LoveMind AI, New York, USA
Diagnosing and Improving Probabilistic Reasoning in Large Language Models
Sep 29, 2026Huaman Sun, Dingcheng Wang, Jason Hartline +1LLM Reasoning Strategies Probabilistic Model
Department of Computer Science Northwestern University Evanston, IL 60201, USA
From Judgment Quality to Downstream Utility: Rethinking LLM-as-a-Judge for Open-Ended Tasks
Sep 29, 2026Zheng Zhang, Lufei Li, Xinyue Tan +4Llm-As-A-Judge Downstream Reasoning
School of Information Science and Technology, ShanghaiTech University · State Key Laboratory of General Artificial Intelligence, BIGAI
Elicitation and Decision Geometry in Single-Index Bandits
Sep 28, 2026Sakshi Arya, Cheng Soon OngMonotonicity Reward Functions
Case Western Reserve University, USA · CSIRO, Australia
How Much Imprecision is Enough Imprecision in my Classifier? A Practical Elicitation Procedure
Sep 27, 2026Victor F. Lopes de Souza, Sébastien Destercke, Abdelhak ImoussatenElicitation Classifier
EuroMov Digital Health in Motion, Univ. Montpellier, IMT Mines Al`es, Montpellier, France · LIRMM, Universit´e de Montpellier, CNRS, Montpellier, France · SyCoIA, IMT Mines Al`es, Al`es, France +1
Persuaded, Not Informed: Incentive-Misaligned Witnesses Defeat In-Context Grounding
Sep 23, 2026Rahul BalakaviPersuasion Elicitation
AmpUp Research
SeeQ: Training Generalist Value Functions for Long-Horizon Robotic Manipulation
Sep 18, 2026Saksham Singh, Zheyuan Hu, Max Sobol Mark +3Robot Policies Robotic Manipulation
Carnegie Mellon University
Corrupt Plans, Clean Traces: Evading Chain-of-Thought Monitoring with Plan Injection
Sep 14, 2026Keertana Chidambaram, Andrew Ilyas, Vasilis SyrgkanisAdversarial Robustness Chain-of-Thought Reasoning
Stanford University · CMU
GUIDE: Generative Utility Inference and Decision Engine
Sep 14, 2026Anagha Tiwari, Alexander G. Gray, Nick Feamster +3Preference Alignment Learning Elicitation
Department of Computer Science, University of Chicago · Centaur AI Institute, University of Chicago · Heinz College & School of Computer Science, Carnegie Mellon University +1
Scored vs. Generated Readouts in Behavioral Language Models: An Empirical Study of Elicitation Format
Sep 9, 2026Touchapon Kraisingkorn, Krittin Pachtrachai, Wachiravit ModecruaFree-Form Textual Rationales Readout
Amity Research and Application Center (ARAC), Amity Group
Navigating the digital spectrum: Assessing political bias, stability, and downstream fairness in Large Language Models
Sep 8, 2026Luka Debevc, Nishan Chatterjee, Antoine Doucet +2Large Language Model Bias Hate Speech Detection
1Jožef Stefan Institute, Ljubljana, Slovenia · University of Ljubljana, Ljubljana, Slovenia · 2Jožef Stefan International Postgraduate School, Ljubljana, Slovenia +1
Marginal Fidelity Does Not Establish User Simulation in Demographic Synthetic Survey Panels: Response Contracts, Support Collapse and Conditioning Failure
Sep 7, 2026Alexander DoudkinSurvey Elicitation
Minds AI Labs, Inc.
LEAP: Likelihood Elicitation and Aggregation for LLM-based Probabilistic Forecasting
Sep 1, 2026Yufei Chen, Yiran Zhao, Xiaogang Xu +3Large Language Model Forecasting Probabilistic Forecasting
1Shandong University · 2Nanjing University of Aeronautics and Astronautics · School of Software Technology, Zhejiang University +2
DualStake: Dual-Path Confidence Calibration in Deep Research Agents
Sep 1, 2026Yinuo Xu, Yuwei Liang, Jianjie Cheng +4Confidence Calibration Deep Research
NLPR & MAIS, Institute of Automation, Chinese Academy of Sciences (CASIA) · School of Artificial Intelligence, University of Chinese Academy of Sciences · Meituan Inc.
BLOOM-WILT: Logit Tilting for Behaviour Elicitation in Automated LLM Auditing
Aug 31, 2026Adrians Skapars, Edoardo ManinoLarge Language Model Evaluation Model Auditing
University of Manchester, United Kingdom
Improving Information Extraction with Learned Queries
Aug 31, 2026Omar Sharif, Soroush Vosoughi, Nikhil SinghInformation Extraction Queries
Department of Computer Science Dartmouth College
R $^2$ A: Learning Persona Policies Through Persona Representation Learning and Runtime Alignment
Aug 30, 2026Mohan Zhang, Chengsong You, Xiaoyu Cao +4Behavioral Foundation Models Artificial Intelligence Alignment
College of AI, Tsinghua University · 2Independent Researcher
ELICITED: EHR-grounded Longitudinal Interactive Conversations for Information-seeking Triage Evaluation and Decision-making
Aug 10, 2026Haohao Zhu, Xiaolin Shi, Jiayu ZhouEmergency Departments Triage
University of Michigan · Ellipsis Health
Scale-to-Dialogue: Low-Burden Elicitation of Daily Premenstrual Symptom Ratings with Small Language Models
Aug 9, 2026Yifan WangElicitation Patients
The Chinese University of Hong Kong, Shenzhen (CUHK-Shenzhen)
Do LLM Recommenders Know When They're Hallucinating? Auditing Confidence Calibration in Catalog Faithfulness
Aug 7, 2026Srijith RavikumarCatalogs Visual Hallucinations
Amazon.com LLC Seattle, Washington, USA
SABRE: Scalable and Automated Benchmarking of VLMs under Stress
Aug 7, 2026Zixuan Lan, Luzhe Sun, Matthew R. Walter +1Vision-Language Foundation Models Double-Edged Sword
University of Chicago · Stony Brook University · Toyota Technological Institute at Chicago
Personalized Deep Research Query Refinement with Graph-Scaffolded Evidence Grounding
Aug 6, 2026Soojin Yoon, Dongha LeeDeep Research Agentic Search
Department of Artificial Intelligence, Yonsei University
AI-assisted Script Management for Requirements Elicitation Interviews
Aug 3, 2026Anmol Singhal, Paulo Carvalho, Travis BreauxInterviews Elicitation
Carnegie Mellon University, Pittsburgh, USA
When Policies Change Probabilities: Modular Decision-Making for LLM Code Review
Aug 2, 2026Rasvik Kudum, Max Corbett, Hitansh Paliwal +3Code Quality Reviewer
DE-NER : Zero-shot Named Entity Recognition via Dialogue Elicitation of Large Language Models
Aug 1, 2026Xuankang Zhang, Jiangming LiuNamed-Entity Recognition Elicitation
Yunnan University
SpecFirst: Behavioral Specification Elicitation as a First-Class Step in Agent-Based Program Synthesis from Scratch
Jul 29, 2026Yihao Chen, Shi Chang, Feng Lin +4Llm-Driven Code Synthesis Coding Agents
2Queen’s University · 1Huawei Canada · 4Concordia University +1
From Dyad to Triad: Eliciting XAI Requirements in Stroke Rehabilitation
Jul 28, 2026Param Rajpura, Yogesh Kumar MeenaRehabilitation Xai
Human-AI Interaction (HAIx) Lab, IIT Gandhinagar, India
Elicitation without Backpropagation: Steering Model Behavior by Optimizing the Latent Posterior
Jul 21, 2026Garrett Baker, Vinayak Pathak, Daniel Murfet +1Posterior Transformer Architectures
Timaeus · CORAL · Monash University
ReqGenX: An Empirical Study of Atomic Decomposition, Artifact Regeneration, and Reconstruction for Legacy SRS Documents
Jul 18, 2026Ragib Shahariar Ayon, Rayed Fahmi, Sumon Biswas +1Empirical Study Elicitation
Texas State University, San Marcos, TX, USA · Case Western Reserve University, Cleveland, OH, USA
Retrieval is Enough: Training-Free Interpretability with a Tool-Using Agent
Jul 17, 2026Sriram Balasubramanian, Soheil FeiziInterpretability Model Activations
Department of Computer Science University of Maryland
PACE: Persona Adaptation through Conversational Elicitation in Human-Robot Interaction
Jul 17, 2026Peizhen Li, Longbing Cao, Megani Rajendran +3Human-Robot Interaction Persona Consistency
Macquarie University, Australia · NVIDIA, Singapore
Evaluating RE Practices for Explainability: Synthesizing Insights from Daimler Truck into an Explainable RE Framework Proposal
Jul 13, 2026Umm-e- Habiba, Lucas Mauser, Jonas Fritzsch +2Explainability Elicitation
TUM School of CIT, Technical University of Munich, Heilbronn, Germany · Daimler Truck AG, Leinfelden-Echterdingen, Germany · Institute of Software Engineering, University of Stuttgart, Stuttgart, Germany +1
NeuReasoner: Theory-grounded Mapping of Reasoning Elicitation Boundaries
Jun 29, 2026Aydin Javadov, Shyngys Aitkazinov, Tobias Hoesli +3Reasoning Skills Elicitation
1ETH Zürich · 2Imperial College London · 3Technical University of Munich
Fuzzing Large Language Models to Elicit Hidden Behaviours
Jun 28, 2026Mohammed Abu Baker, Lakshmi Babu-SaheerFuzzing Elicitation
School of Computing and Information Science Anglia Ruskin University
Probing the Misaligned Thinking Process of Language Models
Jun 23, 2026Kaiwen Zhou, Constantin Venhoff, Jonathan Michala +2Large Language Model Alignment Deception
1Anthropic Fellows · 2MATS · 3UCSC +2
Zero-Shot Active Feature Acquisition via LLM-Elicitation
Jun 17, 2026Binyamin Perets, Natalie Mendelson, Shiran Vainberg +3Pool-Based Active Learning Feature Engineering
Faculty of Medicine, Technion. · NVIDIA
IntElicit: Eliciting and Assessing Contextualized Creativity via Dialogue Policy Optimization
Jun 10, 2026Mingjia Li, Jin Wu, Hong Qian +7Creativity Computerized Adaptive Testing
East China Normal University · Shanghai Innovation Institute
Knowing When to Ask: Self-Gated Clarification for Hierarchical Language Agents
Jun 9, 2026Aijing Gao, Yiming Kang, Mengdie Flora Wang +1Clarification Elicitation
Amazon Web Services
$τ$ -Rec: A Verifiable Benchmark for Agentic Recommender Systems
Jun 8, 2026Bharath Sivaram Narasimhan, Karthik R NarasimhanAgentic Benchmarks Dialogue Benchmarks
Independent Researcher Mountain View, CA, USA · Princeton University Princeton, NJ, USA
Scaffold Effects on GAIA: A Controlled Comparison
Jun 7, 2026Jason StaraceScaffolds Openai
Independent Researcher
A Pre-Registered Causal Partition of Self-Consistency Elicitation and Reward Design in RLVR
Jun 4, 2026Yuze GaoReinforcement Learning With Verifiable Reward Elicitation
A Model of Multi-turn Human Persuadability Using Probabilistic Belief Tracing
Jun 3, 2026Jared Moore, Noah Goodman, Nick Haber +1Persuasion Social Reasoning
Stanford University · University of Washington
Self-Evaluation Is Already There: Eliciting Latent Judge Calibration in Base LLMs with Minimal Data
Jun 3, 2026XiuYu Zhang, Yi Shan, Junfeng Fang +1Self-Evaluation Elicitation
National University of Singapore · Beijing University of Technology
Large Language Models Are Overconfident in Their Own Responses
Jun 2, 2026Mario Sanz-Guerrero, Manuel Mager, Katharina von der WenseInstruction-Tuned Models Overconfidence
∇Johannes Gutenberg University Mainz, Germany · ♠University of Colorado Boulder, USA
On Wednesdays, We Ask Questions: Optimizing "Active Listening" in Automated Legal Triage and Referral
May 29, 2026Quinten Steenhuis, Jacqueline HarveyLegal Reasoning Tasks Legal Domain
Suffolk University Law School Boston, United States of America
Adaptive Interviewing for Persona Simulation in LLMs: Evidence-Grounded Reasoning Improves Decision Alignment
May 28, 2026Ruoxi Su, Yuhan Liu, Jingyu HuUser Simulation Large Language Model Decisions
Independent Researcher · University of Cambridge
Code as a Weapon: A Consensus-Labeled Prompt Bank for Measuring Coding-Model Compliance with Malicious-Code Requests
May 27, 2026Richard J. Young, Gregory D. MoodyCode Quality Large Language Model Safety
University of Nevada Las Vegas · Department of Information Systems · Las Vegas, NV, USA
SNARE: Adaptive Scenario Synthesis for Eliciting Overeager Behavior in Coding Agents
May 27, 2026Yubin Qu, Yi Liu, Gelei Deng +4Coding Agents Synthetic Environments
1Griffith University · 2Quantstamp · 3Nanyang Technological University +2
BC Protocol: Structured Dual-Expert Dialogue for Eliciting High-Quality Chain-of-Thought Post-Training Data
May 25, 2026Bo Zou, Chao XuLLM Reasoning Strategies Elicitation
The Communication Complexity of Instant-Runoff Voting
May 22, 2026Élie de Panafieu, François Durand, Jérôme LangMajority Voting Maximum Independent Set
LAMSADE, CNRS · 1Nokia Bell Labs · 2CNRS, LAMSADE, Universit´e Paris-Dauphine, PSL, France
Smoothed Elicitation Complexity for Approximate $Γ$ -calibration of Discrete Classification Tasks
May 21, 2026Jessica Finocchiaro, Victor Ganson, Drona KhuranaElicitation Approximation
Computer Science, Boston College · Computer Science, University of Colorado Boulder
Efficient Elicitation of Collective Disagreements
May 19, 2026Mohamed Ouaguenouni, Felipe Garrido-Lucero, Umberto Grandi +2Disagreement Majority Voting
IRIT, Université Toulouse Capitole, Toulouse, France · Center for Collective Learning, IAST, Toulouse School of Economics, Toulouse, France · Center for Collective Learning, CIAS, Corvinus University of Budapest, Budapest, Hungary +2
Embedding by Elicitation: Dynamic Representations for Bayesian Optimization of System Prompts
May 18, 2026Zhiyuan Jerry Lin, Benjamin Letham, Samuel Dooley +2Elicitation Prompt Engineering
Meta
Weak-to-Strong Elicitation via Mismatched Wrong Drafts
May 17, 2026Wei DengMismatch Training-Inference Mismatch
Independent Researcher
LLM-Based Persuasion Enables Guardrail Override in Frontier LLMs
May 13, 2026Rodrigo Nogueira, Thales Sales Almeida, Giovana Kerche Bonás +7Argumentation Frameworks Elicitation
Maritaca AI · JusBrasil
The Mechanism of Weak-to-Strong Generalization: Feature Elicitation from Latent Knowledge
May 13, 2026Ryoya Awano, Taiji SuzukiFeature Learning Stochastic Gradient Descent
University of Tokyo · Center for Advanced Intelligence Project, RIKEN
When to Ask a Question: Understanding Communication Strategies in Generative AI Tools
May 11, 2026Charlotte Park, Kate Donahue, Manish RaghavanGenerative Artificial Intelligence Elicitation
Massachusetts Institute of Technology Cambridge, Massachusetts, USA
On Distinguishing Capability Elicitation from Capability Creation in Post-Training: A Free-Energy Perspective
May 8, 2026Yuhao Li, Shengchao LiuPost-Training Supervised Finetuning
Department of Computer Science and Engineering The Chinese University of Hong Kong
Elicitation Matters: How Prompts and Query Protocols Shape LLM Surrogates under Sparse Observations
May 6, 2026Ge Lei, Samuel J. CooperLarge Language Model Uncertainty Elicitation
Dyson School of Design Engineering Imperial College London Exhibition Road, London SW7 2AZ

Sep 30, 2026, cs.AIJ/K move · Enter open · S save

Learning Steganography Is Easy, Learning Steganographic Reasoning Is Hard

Julian Schulz, Lukas Fülle, Rieke Fruengel

Meridian Cambridge · SAIGE

Steganography Chain-of-Thought Reasoning Elicitation

Open paper AskPDF arXiv

Chain-of-thought monitoring as an approach for AI oversight and control is threatened by the possibility of steganographic reasoning, where LLMs conceal their reasoning inside innocuous-looking text. Two neighbouring capabilities, steganographic messaging (passing a concealed message) and encoded reasoning (reasoning in an illegible but unconcealed format), have already been shown to emerge under training pressures that occur in real pipelines, such as reinforcement learning against monitors. This suggests that steganographic reasoning too might arise as an unintended side effect of training. Here, we compare how easily models learn steganographic reasoning and these two neighbouring capabilities across three elicitation methods: reinforcement learning, in-context learning, and supervised fine-tuning (SFT). For most tasks, models learn steganographic reasoning only under SFT, while they learn steganographic messaging and encoded reasoning under all three elicitation methods. Even under SFT, steganographic reasoning requires at least twice as much training as messaging, and for several model-task combinations it is not learned at all. However, on a cover task that makes hiding information especially convenient, steganographic reasoning can be successfully learned under all three elicitation methods. Steganographic reasoning is thus much harder than steganographic messaging and encoded reasoning, and learning the latter two does not imply learning the former. Yet it lies within reach: an easy version is learned under every elicitation method, when the cover task is convenient for hiding information.

Momentum

Related topics

Latest papers 78