DeepScaleR

DeepScaleR is a benchmark dataset used to evaluate the efficiency of reinforcement learning methods, particularly in the context of quantized rollout processes for training large language models.

Papers using DeepScaleR (5)

Reinforcement Learning for Reasoning in Large Language Models with One Training Example2025 · 1 citesQuRL: Efficient Reinforcement Learning with Quantized Rollout2026Prompt Curriculum Learning for Efficient LLM Post-Training2025MATH-Beyond: A Benchmark for RL to Expand Beyond the Base Model2025RL for Reasoning by Adaptively Revealing Rationales2025