6 papers · 1 filter
BOSS: Token-Budgeted Online Sampling and Selection for LLM Reasoning
Yupeng Chang, Yahan Li, Chenlu Guo +3
Test-time scaling can improve LLM reasoning, but existing approaches typically adapt generation or apply answer selection after a predetermined generation schedule. We study within…
CALIBUDGET: Calibration-Guided Source Allocation for Fixed-Budget Mixed-Reasoning Adaptation
Yupeng Chang, Yuan Wu
Fixed-budget adaptation from heterogeneous data sources requires deciding not only how much data to use, but how much exposure each source receives. Size-proportional rules can cro…
Learned Reporting Preferences in RLVR Can Conflict with the Current Request
Yupeng Chang, Wenxuan Zhang, Yuan Wu
Reinforcement learning with verifiable rewards (RLVR) has become a prominent approach for improving language-model performance on reasoning tasks using automatically checked answer…
From Checkpoint Variation to Selection Gains in Supervised Fine-Tuning
Yupeng Chang, Wenxuan Zhang, Yuan Wu
Checkpoint selection is a routine decision in supervised fine-tuning (SFT): training produces multiple checkpoints, but only one is retained. Yet fixed-budget comparisons do not by…
SOS-LoRA: Static Orthogonal-Subspace Low-Rank Adaptation with Fixed Multi-Scale Scaling
Yupeng Chang, Yuan Wu, Yi Chang
Low-Rank Adaptation (LoRA) is a widely used parameter-efficient fine-tuning (PEFT) method for large language models. Under a fixed rank budget, LoRA parameterizes each adapted weig…
BV-Blend: Uncertainty-Weighted Historical Baselines for Stable Critic-Free RL with Verifiable Rewards
Yupeng Chang, Yuan Wu, Yi Chang
Critic-free reinforcement learning with verifiable rewards (RLVR), exemplified by Group Relative Policy Optimization (GRPO), avoids training a value function (critic) and reduces m…