It Takes 8 Tokens: Weak-to-Strong Off-Policy RL via Auxiliary Branches

arXiv:2607.16205v1 Announce Type: new Abstract: Reinforcement learning with verifiable rewards has emerged as a standard approach for enhancing reasoning in large language models, which typically optimizes the policy by contrasting multiple self generated rollouts. However, we identify a critical support limited bottleneck in this paradigm: on challenging reasoning tasks, the target model's samples often exhibit semantic redundancy, converging into the same erroneous "reasoning basins" that offer negligible reward contrast for policy updates. In this paper, we propose to overcome this limitation through a weak to strong learning paradigm, where a policy's exploration is informed by a weaker but computationally efficient auxiliary model. We introduce W2SPO, an off policy RL method that injects short auxiliary segments often as brief as 8 tokens into intermediate target model trajectories and the target model then completes the reasoning path from these diverted states. Policy updates are restricted to these short inserted segments based on final verifiable rewards. Empirically, W2SPO achieves superior performance among evaluated 4B scale models on mathematical reasoning benchmarks, outperforming evaluated post trained baselines. Compared with vanilla GRPO under the same sampling budget, W2SPO improves Pass@1 from 62.3% to 64.2% while achieving a 3.55 times training speedup. These results suggest that weak auxiliary branches can induce stronger target reasoning policies by expanding local exploration support.

Abstract:Reinforcement learning with verifiable rewards has emerged as a standard approach for enhancing reasoning in large language models, which typically optimizes the policy by contrasting multiple self generated rollouts. However, we identify a critical support limited bottleneck in this paradigm: on challenging reasoning tasks, the target model's samples often exhibit semantic redundancy, converging into the same erroneous "reasoning basins" that offer negligible reward contrast for policy updates. In this paper, we propose to overcome this limitation through a weak to strong learning paradigm, where a policy's exploration is informed by a weaker but computationally efficient auxiliary model. We introduce W2SPO, an off policy RL method that injects short auxiliary segments often as brief as 8 tokens into intermediate target model trajectories and the target model then completes the reasoning path from these diverted states. Policy updates are restricted to these short inserted segments based on final verifiable rewards. Empirically, W2SPO achieves superior performance among evaluated 4B scale models on mathematical reasoning benchmarks, outperforming evaluated post trained baselines. Compared with vanilla GRPO under the same sampling budget, W2SPO improves Pass@1 from 62.3% to 64.2% while achieving a 3.55 times training speedup. These results suggest that weak auxiliary branches can induce stronger target reasoning policies by expanding local exploration support.

Artificial Intelligence (cs.AI)

arXiv:2607.16205 [cs.AI]

(or arXiv:2607.16205v1 [cs.AI] for this version)

https://doi.org/10.48550/arXiv.2607.16205 Focus to learn more arXiv-issued DOI via DataCite

arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.

Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.

Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.

aiqbench analysis

It Takes 8 Tokens: Weak-to-Strong Off-Policy RL via Auxiliary Branches

Related coverage