AI & ML
The Post-Training Dilemma: Why We Should Rethink the Sequential SFT-RL Paradigm
arXiv math.IT0 views
arXiv:2601.07389v3 Announce Type: replace-cross Abstract: This position paper argues that the sequential supervised fine-tuning followed by reinforcement learning (SFT-then-RL) pipeline---the de facto paradigm for large language model (LLM) post-training---has a fundamental property that cannot be fixed by hyperparameter tuning: SFT and RL optimize over nearly orthogonal parameter subspaces, so RL inevitably degrades the capabilities acquired during the preceding SFT phase. Under the Polyak-Lojasiewicz (PL) condition, which is empirically observed for over-parameterized networks, we show that RL increases the SFT loss by at least $\frac{\mu}{2}\mathrm{dist}(\theta_{\mathrm{RL}}, \Theta^*_{\mathrm{SFT}})^2$ for some $\mu>0$. Although the KL penalty can constrain distributional shift, it cannot correct the underlying gradient misalignment between the two objectives. Experiments on Qwen3-0.6B across four tasks confirm the theoretical predictions: on MATH-500, Alpaca, HH-RLHF, and SST-2, the SFT loss increases after the RL phase. We identify the underlying mechanism as persistent near-orthogonality between the gradients of the two objectives, paired with opposing landscape dynamics: SFT flattens the loss landscape while RL re-sharpens it. Based on our analysis, we propose three promising research directions: (1) joint optimization of the two objectives with gradient surgery to resolve conflicting updates; (2) landscape-aware scheduling that terminates RL training at the theoretically optimal point; and (3) characterizing the model scale threshold above which SFT becomes unnecessary.
Source:https://arxiv.org/abs/2601.07389Related posts
VESTA: Visual Exploration with Statistical Tool Agents
0arXiv stat.CO
An interpretable semiparametric Bayesian machine learning model for high-dimensional circulating epigenetic signatures of multiple myeloma
0arXiv stat.CO
Quantum Nondecimated Wavelet Transform: Theory, Circuits, and Applications
0arXiv stat.CO
dapper: Data Augmentation for Private Posterior Estimation in R
0arXiv stat.CO
HAVAPlot: A Web-Based Scientific Visualization and Post-Processing Platform
0arXiv stat.CO
ISC Stormcast For Friday, November 20th, 2026 https://isc.sans.edu/podcastdetail/10188, (Fri, Nov 20th)
0SANS ISC