Temporal Leakage in LLM Backtesting: Measurement, Validation, and Adjusted Scores
Zeyu Zhang, Bradly C. Stadie
cs.LG, cs.CL, stat.ML
Submitted: 2026-08-04
Comments: 12 pages main content, 45 pages in total
Code: https://github.com/ZeyuZhang1901/Temporal-Leakage-Backtesting
License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Terminology
Sources
- Look-Ahead-Bench: a Standardized Benchmark of Look-ahead Bias in Point-in-Time LLMs for Finance
- Do Membership Inference Attacks Work on Large Language Models?
- Detecting Lookahead Bias in LLM Forecasts
- Benchmark Inflation: Revealing LLM Performance Gaps Using Retro-Holdouts
- Chronologically Consistent Large Language Models
- LiveCodeBench: Holistic and Contamination Free Evaluation of Large Language Models for Code
- Investigating Data Contamination for Pre-training Language Models
- Summoning the Oracle to Slay It: Mitigating Look-Ahead Bias in Financial Backtesting with Large Language Models
- ExAnte: A Benchmark for Ex-Ante Inference in Large Language Models
- The Memorization Problem: Can We Trust LLMs' Economic Forecasts?
- Proving Test Set Contamination in Black Box Language Models
- Pitfalls in Evaluating Language Model Forecasters
- A Comprehensive Survey of Contamination Detection Methods in Large Language Models
- Quantifying Contamination in Evaluating Code Generation Capabilities of Language Models
- Quantifying the Effect of Test Set Contamination on Generative Evaluations
- Evaluation data contamination in LLMs: how do we measure it and (when) does it matter?
- Hubble: a Model Suite to Advance the Study of LLM Memorization
- LiveBench: A Challenging, Contamination-Limited LLM Benchmark
- Benchmark Data Contamination of Large Language Models: A Survey
- DatedGPT: Preventing Lookahead Bias in Large Language Models with Time-Aware Pretraining
Related papers
- Polynomial-Augmented Neural Networks (PANNs) with Weak Orthogonality Constraints for Enhanced Function and PDE Approximation
- AIRL-S: Unifying Reinforcement Learning and Search-Based Test-Time Scaling via Adversarial Inverse Reinforcement Learning
- Transformers as Bayesian In-Context Experimenters: Smoothness-Adaptive Efficient ATE Estimation
- Convergence issues in Relational Concept Analysis based on AOC-posets
- Beliefs Beyond Posteriors: The Predictive Cost of ELBO Binding
- Understanding Diffusion Models via Ratio-Based Function Approximation with SignReLU Networks