Zeyu (Steven) Zhang

PhD Student, Statistics and Data Science, Northwestern University

I am a third-year Ph.D. student in Statistics and Data Science at Northwestern University, working with Prof. Bradly C. Stadie. My research sits at the intersection of statistics, machine learning, and large language models.

I study how to make LLMs more trustworthy and reliable when they are used for prediction and reasoning—especially failure modes such as hallucination, temporal contamination in evaluation, and explanations that look fluent but may not track the model’s actual decisions. I am particularly interested in interpretation of LLMs and in methods that help us audit, constrain, and improve model behavior under realistic deployment conditions.

Research interests

Trustworthy LLMs Reliability Hallucination Interpretation LLM forecasting

News

Aug 03, 2026	The capstone of our backtesting line is complete: Temporal Leakage in LLM Backtesting: Measurement, Validation, and Adjusted Scores proves the standard contamination check is uninformative and shows one defensible reference restores a leakage-adjusted score. In submission to TMLR, with code and data released.
May 25, 2026	Our new preprint on catching models that “peek” at the future is live on arXiv—with code to reproduce the claim-level leakage audit and TimeSPEC mitigation.
May 13, 2026	TEMPO is out: a training approach that teaches LLMs temporal discipline for trustworthy backtesting. Code released.
Jan 22, 2026	Thrilled that LAMP was selected as a Spotlight at AISTATS 2026—a lightweight way to audit whether a model’s explanations behave like they matter.

Research

Temporal leakage in LLM backtesting

A central challenge in evaluating an LLM’s prediction ability is that testing on the true future is often impractical. The standard alternative is backtesting: ask the model historical questions under a cutoff date, as if it were forecasting from that earlier time. The hard problem is temporal leakage—a special form of data contamination in which post-cutoff knowledge from pretraining seeps into the model’s reasoning and inflates backtest scores. My work treats this problem as a whole. On the evaluation side, a statistical theory shows the standard pre/post contamination check is uninformative—recency alone reproduces the leakage signature—proves that no passive backtest can separate leakage from genuine skill, and shows that one external reference (a known training cutoff or a matched clean control) restores measurement, yielding a validated leakage-adjusted score. On the mitigation side, an inference-time architecture blocks leaked claims from the model’s output without retraining—paired with an interpretable claim-level metric that quantifies how much of the decision-driving reasoning is contaminated—while an RL post-training method teaches the model temporal discipline itself.

Selected works in this line

Temporal leakage causal mechanism and the double-benefit law

Measure & correct the score Leakage Audit

Temporal Leakage in LLM Backtesting: Measurement, Validation, and Adjusted Scores

The standard pre/post contamination check is uninformative: recency alone reproduces the leakage signature, and no passive backtest can separate leakage from skill and recency. One external reference—a known training cutoff or a matched clean control—restores measurement and yields a leakage-adjusted score, validated by planting leakage in twin models and deployed as an audit of frontier systems.

Preprint · In submission to TMLR

Measure & mitigate at inference TimeSPEC

All Leaks Count, Some Count More

Backtesting assumes models use only pre-cutoff evidence, but pretrained LLMs routinely leak later knowledge into their rationales. We introduce Shapley-DCLR to measure how much decision-critical reasoning is contaminated at the claim level, and TimeSPEC, an inference-time supervisor that regenerates predictions from temporally filtered evidence without retraining.

Preprint · Under review at EMNLP 2026

Mitigate via training TEMPO

Temporal Enforcement via Mode-Separated Policy Optimization

Temporal leakage lets LLMs recall post-cutoff outcomes during backtesting, so reported accuracy is often invalid. TEMPO trains temporal discipline with a two-mode RL objective that first drives leaked claims to zero, then optimizes prediction quality once reasoning stays pre-cutoff.

Preprint · Under review at NeurIPS 2026

Interpretation: do explanations match decisions?

LLMs often give fluent explanations for their predictions, but fluency is not faithfulness. The hard question is whether the factors a model cites actually behave like they drive the decision—and how to check that without gradients or internal access. That makes interpretation both practically needed for trust and scientifically interesting as a window into model behavior.

Selected works in this line

Local decision surfaces LAMP

Extracting Locally Linear Decision Surfaces from LLM World Models

Fluent LLM explanations need not track what actually drives the prediction. LAMP treats the model’s own stated factors as coordinates and fits a local linear decision surface—without gradients or internal access—to test whether predictions move consistently when those factors change.

Spotlight paper in AISTATS 2026