# Why Has Predicting Downstream Capabilities of Frontier AI Models with Scale Remained Elusive?

arXiv: https://arxiv.org/abs/2406.04391 · HTML: https://mlsanity.com/paper/2406.04391/

## Key facts

- Cited by 40 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 6 Jun 2024.

## Built on by (most cited first, 30 of 40)

- [Safetywashing: Do AI Safety Benchmarks Actually Measure Safety Progress?](https://mlsanity.com/paper/2407.21792/) (2024)
- [Quantifying Variance in Evaluation Benchmarks](https://mlsanity.com/paper/2406.10229/) (2024)
- [How Do Large Language Monkeys Get Their Power (Laws)?](https://mlsanity.com/paper/2502.17578/) (2025)
- [Emergent Abilities in Large Language Models: A Survey](https://mlsanity.com/paper/2503.05788/) (2025)
- [A Rosetta Stone for AI Benchmarks](https://mlsanity.com/paper/2512.00193/) (2025)
- [Establishing Task Scaling Laws via Compute-Efficient Model Ladders](https://mlsanity.com/paper/2412.04403/) (2024)
- [Training on the Test Task Confounds Evaluation and Emergence](https://mlsanity.com/paper/2407.07890/) (2024)
- [Scaling Laws Are Unreliable for Downstream Tasks: A Reality Check](https://mlsanity.com/paper/2507.00885/) (2025)
- [Signal and Noise: A Framework for Reducing Uncertainty in Language Model Evaluation](https://mlsanity.com/paper/2508.13144/) (2025)
- [Language Models Improve When Pretraining Data Matches Target Tasks](https://mlsanity.com/paper/2507.12466/) (2025)
- [On the Limitations of Compute Thresholds as a Governance Strategy](https://mlsanity.com/paper/2407.05694/) (2024)
- [Loss-to-Loss Prediction: Scaling Laws for All Datasets](https://mlsanity.com/paper/2411.12925/) (2024)
- [Revisiting the Superficial Alignment Hypothesis](https://mlsanity.com/paper/2410.03717/) (2024)
- [Putnam-AXIOM: A Functional and Static Benchmark for Measuring Higher Level Mathematical Reasoning in LLMs](https://mlsanity.com/paper/2508.08292/) (2025)
- [U-shaped and Inverted-U Scaling behind Emergent Abilities of Large Language Models](https://mlsanity.com/paper/2410.01692/) (2024)
- [Optimal Self-Consistency for Efficient Reasoning with Large Language Models](https://mlsanity.com/paper/2511.12309/) (2025)
- [Towards an Improved Understanding and Utilization of Maximum Manifold Capacity Representations](https://mlsanity.com/paper/2406.09366/) (2024)
- [Position: Machine Learning Conferences Should Establish a "Refutations and Critiques" Track](https://mlsanity.com/paper/2506.19882/) (2025)
- [Relative Scaling Laws for LLMs](https://mlsanity.com/paper/2510.24626/) (2025)
- [Revisiting the Scaling Properties of Downstream Metrics in Large Language Model Training](https://mlsanity.com/paper/2512.08894/) (2025)
- [Pretraining Scaling Laws for Generative Evaluations of Language Models](https://mlsanity.com/paper/2509.24012/) (2025)
- [Uncovering Latent Memories: Assessing Data Leakage and Memorization Patterns in Frontier AI Models](https://mlsanity.com/paper/2406.14549/) (2024)
- [Task-Stratified Knowledge Scaling Laws for Post-Training Quantized Large Language Models](https://mlsanity.com/paper/2508.18609/) (2025)
- [Psychometric Personality Shaping Modulates Capabilities and Safety in Language Models](https://mlsanity.com/paper/2509.16332/) (2025)
- [Correlating and Predicting Human Evaluations of Language Models from Natural Language Processing Benchmarks](https://mlsanity.com/paper/2502.18339/) (2025)
- [Evaluating the Robustness of Chinchilla Compute-Optimal Scaling](https://mlsanity.com/paper/2509.23963/) (2025)
- [Predicting Task Performance with Context-aware Scaling Laws](https://mlsanity.com/paper/2510.14919/) (2025)
- [Quantifying the Effect of Test Set Contamination on Generative Evaluations](https://mlsanity.com/paper/2601.04301/) (2026)
- [Neural Neural Scaling Laws](https://mlsanity.com/paper/2601.19831/) (2026)
- [Item Response Scaling Laws: A Measurement Theory Approach for Efficient and Generalizable Neural Scaling Estimation](https://mlsanity.com/paper/2606.07616/) (2026)

## Abstract (arXiv)

Predicting changes from scaling advanced AI systems is a desirable property for engineers, economists, governments and industry alike, and, while a well-established literature exists on how pretraining performance scales, predictable scaling behavior on downstream capabilities remains elusive. While many factors are certainly responsible, this paper identifies a significant factor that makes predicting scaling behavior on widely used multiple-choice question answering benchmarks challenging and illuminates a path towards making such downstream evaluations predictable with scale. Using five model families and twelve well-established multiple-choice benchmarks, we demonstrate that downstream performance is computed from negative log likelihoods via a sequence of transformations that progressively degrades the statistical relationship between performance and scale. We then pinpoint the mechanism causing this degradation: downstream metrics require comparing the correct choice against a small number of specific incorrect choices, meaning accurately predicting downstream capabilities requires predicting not just how probability mass concentrates on the correct choice with scale, but also how probability mass fluctuates on the alternative incorrect choices with scale. We empirically study how probability mass on the correct choice co-varies with probability mass on incorrect choices with increasing compute, suggesting that scaling laws for textitincorrect choices might be achievable. Our work also explains why pretraining scaling laws are commonly regarded as more predictable than downstream capabilities and contributes towards establishing scaling-predictable evaluations of frontier AI models.