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

Rylan Schaeffer, Hailey Schoelkopf, Brando Miranda, Gabriel Mukobi, Varun Madan, Adam Ibrahim, Herbie Bradley, Stella Biderman, Sanmi Koyejo
Proceedings of the 42nd International Conference on Machine Learning, PMLR 267:53177-53266, 2025.

Abstract

Predictable behavior from scaling advanced AI systems is an extremely desirable property for engineers, companies, economists and governments 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 shines a light on 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 show 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 reveal 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 specific 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 incorrect 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.

Cite this Paper

BibTeX
@InProceedings{pmlr-v267-schaeffer25b, title = {Why Has Predicting Downstream Capabilities of Frontier {AI} Models with Scale Remained Elusive?}, author = {Schaeffer, Rylan and Schoelkopf, Hailey and Miranda, Brando and Mukobi, Gabriel and Madan, Varun and Ibrahim, Adam and Bradley, Herbie and Biderman, Stella and Koyejo, Sanmi}, booktitle = {Proceedings of the 42nd International Conference on Machine Learning}, pages = {53177--53266}, year = {2025}, editor = {Singh, Aarti and Fazel, Maryam and Hsu, Daniel and Lacoste-Julien, Simon and Berkenkamp, Felix and Maharaj, Tegan and Wagstaff, Kiri and Zhu, Jerry}, volume = {267}, series = {Proceedings of Machine Learning Research}, month = {13--19 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v267/main/assets/schaeffer25b/schaeffer25b.pdf}, url = {https://proceedings.mlr.press/v267/schaeffer25b.html}, abstract = {Predictable behavior from scaling advanced AI systems is an extremely desirable property for engineers, companies, economists and governments 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 shines a light on 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 show 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 reveal 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 specific 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 incorrect 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.} }
Endnote
%0 Conference Paper %T Why Has Predicting Downstream Capabilities of Frontier AI Models with Scale Remained Elusive? %A Rylan Schaeffer %A Hailey Schoelkopf %A Brando Miranda %A Gabriel Mukobi %A Varun Madan %A Adam Ibrahim %A Herbie Bradley %A Stella Biderman %A Sanmi Koyejo %B Proceedings of the 42nd International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2025 %E Aarti Singh %E Maryam Fazel %E Daniel Hsu %E Simon Lacoste-Julien %E Felix Berkenkamp %E Tegan Maharaj %E Kiri Wagstaff %E Jerry Zhu %F pmlr-v267-schaeffer25b %I PMLR %P 53177--53266 %U https://proceedings.mlr.press/v267/schaeffer25b.html %V 267 %X Predictable behavior from scaling advanced AI systems is an extremely desirable property for engineers, companies, economists and governments 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 shines a light on 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 show 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 reveal 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 specific 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 incorrect 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.
APA
Schaeffer, R., Schoelkopf, H., Miranda, B., Mukobi, G., Madan, V., Ibrahim, A., Bradley, H., Biderman, S. & Koyejo, S.. (2025). Why Has Predicting Downstream Capabilities of Frontier AI Models with Scale Remained Elusive?. Proceedings of the 42nd International Conference on Machine Learning, in Proceedings of Machine Learning Research 267:53177-53266 Available from https://proceedings.mlr.press/v267/schaeffer25b.html.

Related Material