Poster Session 2

Posters scheduled for this session have it as their primary poster session. Presenters may also join the other session if space permits.

Poster Session 1: 10:45–12:00 PDT. Poster Session 2: 15:00–16:15 PDT.

October 9, 2026 · Workshop room 10: Franciscan A&B / Ballroom level.

Scheduling note: Xiaoxue Yang, presenting “Auditing Prompt Injection Defenses Before They Reach Users: What Strong Adversaries Reveal,” has requested to begin at 15:45 PDT.

Assigned posters (39)

25 have confirmed in-person attendance; 14 have a default session assignment with attendance unconfirmed. Authors with a scheduling conflict may request a change.

• ACID-Bench: Measurement Contracts for State-Changing Tool Agents — Default assignment; attendance unconfirmed.
• Adaptive AI Evaluation via Entropy-Stopped Bayesian Probing of Recurrent Agents — Default assignment; attendance unconfirmed.
• Algorithmic Blindness in Large Language Models: A Calibration Study of Performance Prediction
• Auditing Prompt Injection Defenses Before They Reach Users: What Strong Adversaries Reveal
• Can LLMs Predict the Future? PolyBench: A Benchmark for Future Prediction from Real- World Markets — Default assignment; attendance unconfirmed.
• Code Migration Benchmark: Evaluating AI Models' Autonomous Code Migration Performance
• Do LLM-Elicited Correlation Matrices Exist? A Multi-Model Positive-Semidefiniteness Probe — Default assignment; attendance unconfirmed.
• Does the Proxy Preserve the Decision? A Reference-Sensitivity Audit of Multiple-Choice Reformulations of Incremental QA
• Evaluating Generative Agents with Actions Grounded in Socially Distributed Task Environments using Incognita
• Growing Pains: Extensible and Efficient LLM Benchmarking Via Fixed Parameter Calibration — Default assignment; attendance unconfirmed.
• Legal Research Bench: Measuring End-to-End Reliability in Long-Horizon Legal Research Agents
• LLM-as-a-Verifier: A General-Purpose Verification Framework
• Measuring Decision-Scale Use in Tool-Augmented LLMs: A Contrastive Urban Benchmark
• Measuring Inter-Agent Trust: A Memoryless-Anchor Instrument for Interactive, Non-Stationary Evaluation
• Measuring Prompt-Injection Defenses for Computer-Use Agents: Silent Failures, Landed Attacks, and the Block– Confirm Frontier — Default assignment; attendance unconfirmed.
• Measuring Reward Hacking and Reasoning-Answer Decoupling Under Position-Confounded Optimization
• Measuring the Measurers: A Generalizability Theory Approach to LLM-as-a-Judge Reliability
• Measuring Whether Black-Box LLM APIs Know When They Might Be Wrong — Default assignment; attendance unconfirmed.
• Minimal Execution Patches for Causal Debugging of Tool-Using Agents — Default assignment; attendance unconfirmed.
• PhantomBench: Benchmarking the Non-existential Threat of Language Models
• PhantomFill: When the Form Demands an Answer, Language Models Invent One — Default assignment; attendance unconfirmed.
• Preserving What Matters: Semantic Scaffolds Beyond Saturation in Summarization Evaluation
• Public Benefits Bench: Can AI Help People Navigate SNAP Benefits?
• Rating the Raters: Rasch Measurement Theory for LLM Evaluation
• Replayable Trajectory Dataflow: Auditing Reward Measurement under Optimization and Drift in RLVR
• Rethinking Psychometric Evaluation of LLMs: When and Why Self-Reports Predict Behavior
• Skill-based Agentic Evaluation for Real-time Data Science Tasks
• Spiking the training data to correct for test set contamination
• Statistical Foundations of LLM-based A/B Testing: A Surrogacy Framework for Human Causal Inference
• Stochastic Collapse: Video World Models Fail to Reproduce Outcome Frequencies
• The Castle of Crossed Models: Benchmarking LLMs on Calvinian Narrative Crossings
• The dstate Measurement Problem: A Construct-Validity and Non-Stationarity Audit of Recurrent-State Capacity in 20 Linear Architectures — Default assignment; attendance unconfirmed.
• The Lifecycle of LLM-as-a-Judge for Large-Scale Recommendation Explanations
• TRACE: Diagnosing Verifier Brittleness in Agentic Evaluation
• UDET: A Dataset for Detecting Contamination on Competition Math via Underdetermined Problems — Default assignment; attendance unconfirmed.
• Verbalized Confidence Across Autoregressive and Diffusion Language Models
• WebsiteBench: Browser-Evidence Reconstruction for Website-Cloning Agents — Default assignment; attendance unconfirmed.
• When Does a Probe Still Measure Its Construct? Construct Validity of Eval-Awareness Probes under Non-Stationarity (RLVR) — Default assignment; attendance unconfirmed.
• When Model and Judge Are Both AI: Decomposing the Reliability of Open-Ended Math Grading — Default assignment; attendance unconfirmed.