13 papers
When Are Aggregate Agent Traces Diagnosable? Traffic-Governed Interpretation and Calibrated Abstention
Peiying Zhu, Sidi Chang
Runtime traces can appear transparent, but a closed-loop policy determines which states are visited and which failures become visible. We study a simulated hotel-pricing agent mapp…
Refuse, Decompose, Refresh: A Claim-Safe Protocol for Closed-Loop AI Evaluation
Peiying Zhu, Sidi Chang
An AI evaluation can be perfectly reproducible and still support the wrong claim. This risk is acute in closed-loop systems: policy determines visited states, observable components…
Four Ledgers, Not One Score: Responsible Communication of LLM-Judge Calibration in Biomedical ML
Sidi Chang, Peiying Zhu
Synthetic perturbations appear to offer inexpensive calibration data for LLM evaluators in biomedical ML, where expert review is scarce. Yet a planted mutation key is neither a det…
A Verifier Can Leak the Answer: Diagnosability Before Optimization in Closed-Loop Agent Debugging
Peiying Zhu, Sidi Chang
Agent developers increasingly compare prompts, tools, policies, and diagnosis algorithms through simulator-grounded verifiers. A verifier can nevertheless make a solver comparison…
High-Value Synthetic Supervision for Parameter-Efficient Adaptation of a Compact Japanese Speech Model
Sidi Chang, Peiying Zhu
Private domain speech is difficult to collect and redistribute, while compact models need task-specific supervision. We study an auditable synthetic pipeline that maps Japanese car…
ClaimReceipt: Verifying Evidence Sufficiency and Coverage in Agent Evaluations
Peiying Zhu, Sidi Chang
Agent evaluations face two distinct evidentiary questions: whether a reported claim is recomputable from retained evidence (sufficiency), and whether the retained records cover the…
When Guardrails Look Effective: Construct Validity Failures in LLM Agent Commerce Evaluation
Peiying Zhu, Sidi Chang
Interactive simulations increasingly evaluate policies in markets populated by language-model agents. Their outputs can look economic---prices, profits, consumer surplus, and welfa…
When Aggregate Alignment Misleads: Auditing Policy Repair Without Per-State Expert Actions
Peiying Zhu, Sidi Chang
Agentic AI systems are increasingly used to edit, refine, and repair decision policies, but evaluating these edits is difficult when per-state expert action labels are unavailable.…
Market-Alignment Risk in Pricing Agents: Trace Diagnostics and Trace-Prior RL under Hidden Competitor State
Peiying Zhu, Sidi Chang
Outcome metrics can certify the wrong behavior. We study this failure in a two-hotel revenue-management simulator where Hotel A trains an agent against a fixed rule-based revenue-m…
ValueBlindBench: Agreement-Gated Stress Testing of LLM-Judged Investment Rationales Before Returns Are Observable
Sidi Chang, Peiying Zhu, Yuxiao Chen
LLM-based financial agents increasingly produce investment rationales before the outcomes needed to evaluate them are observable. This creates a delayed-ground-truth evaluation pro…
When Outcome Looks Right But Discipline Fails: Trace-Based Evaluation Under Hidden Competitor State
Peiying Zhu, Sidi Chang
Outcome-only evaluation can certify economically unsafe agents: a policy can hit a business KPI while violating deployable behavioral discipline. In hotel pricing with hidden compe…
Measurement Risk in Supervised Financial NLP: Rubric and Metric Sensitivity on JF-ICR
Sidi Chang, Peiying Zhu, Yuxiao Chen +1
As LLMs become credible readers of earnings calls, investor-relations Q\&A, guidance, and disclosure language, supervised financial NLP benchmarks increasingly function as decision…
The End of Rented Discovery: How AI Search Redistributes Power Between Hotels and Intermediaries
Peiying Zhu, Sidi Chang
When a traveler asks an AI search engine to recommend a hotel, which sources get cited -- and does query framing matter? We audit 1,357 grounding citations from Google Gemini acros…