Close Menu
techwavedispatch.com

    Subscribe to Updates

    Get the latest creative news from FooBar about art, design and business.

    What's Hot

    Internet Archive Staffers Are Unionizing

    November 20, 2026

    Improving Imitation Policies for Multi-Phase Tasks through Residual Learning Allocation

    November 20, 2026

    Nothing’s AI Device Debuts December 8, And It Looks Like Two Vapes Stuck Together

    November 20, 2026
    Facebook X (Twitter) Instagram
    Trending
    • Internet Archive Staffers Are Unionizing
    • Improving Imitation Policies for Multi-Phase Tasks through Residual Learning Allocation
    • Nothing’s AI Device Debuts December 8, And It Looks Like Two Vapes Stuck Together
    • Why We Should Rethink the Sequential SFT-RL Paradigm
    • Chips can survive the space environment. Can their financing?
    • Prescribed Bayesian Efficiency under Lifetime Adjustment Costs
    • Robot Videos: Morphing Robots, Dancing Humanoids, More
    • You Can Reveal The Full GTA 6 Map In Minutes Using This Simple Plane Trick
    Facebook X (Twitter) Instagram
    techwavedispatch.com
    Subscribe
    Friday, November 20
    • Home
    • AI Tools & Software
    • Machine Learning
    • Generative AI
    • Robotics & Automation
    • Future Tech & Hardware
    • AI in Business
    • Ethics & Regulations
    • More
      • Gadgets & Smart Devices
      • Cybersecurity & AI
      • Tech Breakthroughs
    techwavedispatch.com
    Home»Machine Learning»Why We Should Rethink the Sequential SFT-RL Paradigm
    Machine Learning

    Why We Should Rethink the Sequential SFT-RL Paradigm

    adminBy adminNovember 20, 2026No Comments2 Mins Read
    Facebook Twitter Pinterest LinkedIn Tumblr Email
    Why We Should Rethink the Sequential SFT-RL Paradigm
    Share
    Facebook Twitter LinkedIn Pinterest Email

    [Submitted on 12 Jan 2026 (v1), last revised 19 Nov 2026 (this version, v3)]

    View a PDF of the paper titled The Post-Training Dilemma: Why We Should Rethink the Sequential SFT-RL Paradigm, by Xueyan Niu and 3 other authors

    View PDF
    HTML (experimental)

    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.

    Submission history

    From: Xueyan Niu [view email]
    [v1]
    Mon, 12 Jan 2026 10:14:09 UTC (170 KB)
    [v2]
    Wed, 6 May 2026 07:45:21 UTC (176 KB)
    [v3]
    Thu, 19 Nov 2026 01:42:25 UTC (105 KB)

    Paradigm rethink Sequential SFTRL
    Share. Facebook Twitter Pinterest LinkedIn Tumblr Email
    admin
    • Website

    Related Posts

    Improving Imitation Policies for Multi-Phase Tasks through Residual Learning Allocation

    November 20, 2026

    Prescribed Bayesian Efficiency under Lifetime Adjustment Costs

    November 20, 2026

    State-Space Compression for AI-Based Radio Resource Management of Vehicular Services

    November 20, 2026
    Leave A Reply Cancel Reply

    Top Posts

    NVIDIA Is Investigating Reports That CEO Jensen Huang Was The Target Of Chinese Spy Efforts

    November 8, 202619 Views

    ADIndex: Layer-Adaptive Sparse Attention Indexing via Group-Level Distillation

    November 11, 20265 Views

    Anthropic releases new Mythos and Fable models amid RSI concerns

    November 9, 20265 Views

    If AGI is so close, why is everyone leaving? The neolabs and their bets

    November 6, 20265 Views
    Don't Miss
    AI in Business

    Internet Archive Staffers Are Unionizing

    By adminNovember 20, 20260

    Staffers at the Internet Archive, the non-profit “digital library” that runs the Wayback Machine, have…

    Improving Imitation Policies for Multi-Phase Tasks through Residual Learning Allocation

    November 20, 2026

    Nothing’s AI Device Debuts December 8, And It Looks Like Two Vapes Stuck Together

    November 20, 2026

    Why We Should Rethink the Sequential SFT-RL Paradigm

    November 20, 2026
    About Us

    Welcome to TechWave Dispatch

    At TechWave Dispatch, we bring you the latest news, insights, and developments from the ever-evolving world of technology. Our mission is to keep readers informed about the innovations shaping our digital future.

    From artificial intelligence and machine learning to cybersecurity, robotics, smart devices, and groundbreaking technologies, we cover the topics that matter in today's fast-moving tech landscape.

    Our Picks

    Internet Archive Staffers Are Unionizing

    November 20, 2026

    Improving Imitation Policies for Multi-Phase Tasks through Residual Learning Allocation

    November 20, 2026

    Nothing’s AI Device Debuts December 8, And It Looks Like Two Vapes Stuck Together

    November 20, 2026
    Most Popular

    After blocking Muse, Amazon now lets users text Alexa to help them shop

    November 6, 20260 Views

    A controlled world for robot learning and evaluation

    November 6, 20260 Views

    Perplexity now has its own always-on AI agent

    November 6, 20260 Views
    Facebook X (Twitter) Instagram Pinterest
    • Terms and Conditions
    • Privacy Policy
    • Contact Us
    • About Us
    © 2026 TechWave Dispatch. All Rights Reserved.

    Type above and press Enter to search. Press Esc to cancel.