Keep Guessing? When Considering Inference Scaling, Mind the Baselines

• URL: http://arxiv.org/abs/2410.15466v1
• Date: Sun, 20 Oct 2024 18:43:05 GMT
• Title: Keep Guessing? When Considering Inference Scaling, Mind the Baselines
• Authors: Gal Yona, Or Honovich, Omer Levy, Roee Aharoni,
• Abstract summary: Scaling inference compute in large language models consistently increases the coverage (fraction of problems solved) as the number of samples increases. We conjecture that this observed improvement is partially due to the answer distribution of standard evaluation benchmarks, which is skewed towards a relatively small set of common answers.
• Score: 45.21178011740911
• License:
• Abstract: Scaling inference compute in large language models (LLMs) through repeated sampling consistently increases the coverage (fraction of problems solved) as the number of samples increases. We conjecture that this observed improvement is partially due to the answer distribution of standard evaluation benchmarks, which is skewed towards a relatively small set of common answers. To test this conjecture, we define a baseline that enumerates answers according to their prevalence in the training set. Experiments spanning two domains -- mathematical reasoning and factual knowledge -- reveal that this baseline outperforms repeated model sampling for some LLMs, while the coverage for others is on par with that of a mixture strategy that obtains $k$ answers by using only $10$ model samples and similarly guessing the remaining $k-10$ attempts via enumeration. Our baseline enables a more accurate measurement of how much repeated sampling improves coverage in such settings beyond prompt-agnostic guessing.

Related papers

• Large Language Monkeys: Scaling Inference Compute with Repeated Sampling [81.34900892130929]
  We explore inference compute as another axis for scaling by increasing the number of generated samples. In domains like coding and formal proofs, where all answers can be automatically verified, these increases in coverage directly translate into improved performance. We find that identifying correct samples out of many generations remains an important direction for future research in domains without automatic verifiers.
  arXiv  Detail & Related papers  (2024-07-31T17:57:25Z)
• Regression-aware Inference with LLMs [52.764328080398805]
  We show that an inference strategy can be sub-optimal for common regression and scoring evaluation metrics. We propose alternate inference strategies that estimate the Bayes-optimal solution for regression and scoring metrics in closed-form from sampled responses.
  arXiv  Detail & Related papers  (2024-03-07T03:24:34Z)
• Collapsed Inference for Bayesian Deep Learning [36.1725075097107]
  We introduce a novel collapsed inference scheme that performs Bayesian model averaging using collapsed samples. A collapsed sample represents uncountably many models drawn from the approximate posterior. Our proposed use of collapsed samples achieves a balance between scalability and accuracy.
  arXiv  Detail & Related papers  (2023-06-16T08:34:42Z)
• Saliency Grafting: Innocuous Attribution-Guided Mixup with Calibrated Label Mixing [104.630875328668]
  Mixup scheme suggests mixing a pair of samples to create an augmented training sample. We present a novel, yet simple Mixup-variant that captures the best of both worlds.
  arXiv  Detail & Related papers  (2021-12-16T11:27:48Z)
• Learning with Comparison Feedback: Online Estimation of Sample Statistics [2.7158841992922875]
  We study an online version of the noisy binary search problem where feedback is generated by a non-stochastic rather than perturbed by random noise. We maintain an accurate estimate for the median of an adversary adversarial sequence of integers.
  arXiv  Detail & Related papers  (2021-01-11T20:28:32Z)
• Optimal Off-Policy Evaluation from Multiple Logging Policies [77.62012545592233]
  We study off-policy evaluation from multiple logging policies, each generating a dataset of fixed size, i.e., stratified sampling. We find the OPE estimator for multiple loggers with minimum variance for any instance, i.e., the efficient one.
  arXiv  Detail & Related papers  (2020-10-21T13:43:48Z)
• Multinomial Sampling for Hierarchical Change-Point Detection [0.0]
  We propose a multinomial sampling methodology that improves the detection rate and reduces the delay. Our experiments show results that outperform the baseline method and we also provide an example oriented to a human behavior study.
  arXiv  Detail & Related papers  (2020-07-24T09:18:17Z)
• Multi-label Contrastive Predictive Coding [125.03510235962095]
  Variational mutual information (MI) estimators are widely used in unsupervised representation learning methods such as contrastive predictive coding (CPC) We introduce a novel estimator based on a multi-label classification problem, where the critic needs to jointly identify multiple positive samples at the same time. We show that using the same amount of negative samples, multi-label CPC is able to exceed the $log m$ bound, while still being a valid lower bound of mutual information.
  arXiv  Detail & Related papers  (2020-07-20T02:46:21Z)
• Learning Entangled Single-Sample Distributions via Iterative Trimming [28.839136703139225]
  We analyze a simple and computationally efficient method based on iteratively trimming samples and re-estimating the parameter on the trimmed sample set. We show that the method in logarithmic iterations outputs an estimation whose error only depends on the noise level of the $lceil alpha n rceil$-th noisiest data point.
  arXiv  Detail & Related papers  (2020-04-20T18:37:43Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.