Pith. sign in

REVIEW 12 cited by

Supervised Fine Tuning on Curated Data is Reinforcement Learning (and can be improved)

Not yet reviewed by Pith; the record is open.

This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.

SPECIMEN: schema-true, not a live event

T0 review · schema-true

One-sentence machine reading of the paper's core claim.

pith:XXXXXXXX · record.json · timestamp

arxiv 2507.12856 v2 pith:J4ZCC52W submitted 2025-07-17 cs.LG cs.AI

Supervised Fine Tuning on Curated Data is Reinforcement Learning (and can be improved)

classification cs.LG cs.AI
keywords datacuratedlearningpoliciessupervisedtrainingboundcontrol
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved
0 comments
read the original abstract

Behavior Cloning (BC) on curated (or filtered) data is the predominant paradigm for supervised fine-tuning (SFT) of large language models; as well as for imitation learning of control policies. Here, we draw on a connection between this successful strategy and the theory and practice of finding optimal policies via Reinforcement Learning (RL). Building on existing literature, we clarify that SFT can be understood as maximizing a lower bound on the RL objective in a sparse reward setting. Giving support to its often observed good performance. From this viewpoint, we realize that a small modification to SFT leads to an importance weighted variant that behaves closer to training with RL as it: i) optimizes a tighter bound to the RL objective and, ii) can improve performance compared to SFT on curated data. We refer to this variant as importance weighted supervised fine-tuning (iw-SFT). We show that it is easy to implement and can be further generalized to training with quality scored data. The resulting SFT variants are competitive with more advanced RL algorithms for large language models and for training policies in continuous control tasks. For example achieving 66.7% on the AIME 2024 dataset.

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Forward citations

Cited by 12 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. Beyond Trajectory Imitation: Strategy-Guided Policy Optimization for LLM Reasoning

    cs.AI 2026-06 unverdicted novelty 7.0

    SGPO extracts strategies from strong-model responses, builds autonomous and guided trajectories, and applies token-level forward-KL distillation with adaptive weighting to outperform SFT and RL baselines by 2.2 points...

  2. Reference-Sampled Boltzmann Projection for KL-Regularized RLVR: Target-Matched Weighted SFT, Finite One-Shot Gaps, and Policy Mirror Descent

    cs.LG 2026-05 unverdicted novelty 7.0

    Reference-sampled weighted SFT with prompt-normalized Boltzmann weights induces the same policy as fixed-reference KL-regularized RLVR, with BOLT as the estimator and a finite one-shot error decomposition separating c...

  3. DRIFT: Decoupled Rollouts and Importance-Weighted Fine-Tuning for Efficient Multi-Turn Optimization

    cs.LG 2026-05 unverdicted novelty 6.0

    DRIFT achieves multi-turn RL performance via offline importance-weighted SFT by leveraging the equivalence of KL-regularized RL to weighted supervised learning.

  4. Rotation-Preserving Supervised Fine-Tuning

    cs.LG 2026-05 unverdicted novelty 6.0

    RPSFT improves the in-domain versus out-of-domain performance trade-off during LLM supervised fine-tuning by penalizing rotations in pretrained singular subspaces as a proxy for loss-sensitive directions.

  5. Decouple before Integration: Test-time Synthesis of SFT and RLVR Task Vectors

    cs.LG 2026-05 conditional novelty 6.0

    DoTS decouples SFT and RLVR training then synthesizes their task vectors at inference time to match integrated training results at ~3% compute cost.

  6. Multi-LLM Orchestration for High-Quality Code Generation: Exploiting Complementary Model Strengths

    cs.SE 2025-10 conditional novelty 6.0

    PerfOrch is a four-agent multi-LLM system that uses offline profiling to build language-and-category rankings for routing tasks, achieving 97.19% and 95.83% pass@1 on HumanEval-X and EffiBench-X with generalization ac...

  7. A Few Teacher Steps Go a Long Way: Cost-Efficient On-Policy Data Augmentation for Agent Post-Training

    cs.LG 2026-07 conditional novelty 5.5

    A few unfiltered teacher turns at learner-induced contexts are a more cost-efficient SFT allocation than longer or success-filtered teacher completions for LLM agents.

  8. PriFT: Prior-Support Guided Supervised Fine-Tuning

    cs.CL 2026-06 unverdicted novelty 5.0

    PriFT uses token reweighting signals from a frozen pretrained model to stabilize SFT and achieve better results than standard SFT baselines on reasoning tasks.

  9. When RL Fails after SFT: Rejuvenating Model Plasticity for Robust SFT-to-RL Handoff

    cs.LG 2026-06 unverdicted novelty 5.0

    Excessive SFT reduces LLM plasticity for RL; Rejuvenation restores it via base-anchored fusion and targeted neuron resets, yielding better RL performance and OOD generalization.

  10. Proximal Supervised Fine-Tuning

    cs.LG 2025-08 unverdicted novelty 5.0

    PSFT modifies supervised fine-tuning by incorporating trust-region ideas from RL to constrain policy changes, yielding better out-of-domain generalization in math and human-value tasks without entropy collapse.

  11. Entropy-KL Divergence-based Token Masking: A Novel Approach for Selective Fine-tuning of Large Language Models

    cs.AI 2026-05 unverdicted novelty 4.0

    EKSFT masks high-entropy or high-KL tokens in low-data SFT to preserve pre-trained distribution and improve downstream RL performance on math reasoning tasks.

  12. Compatibility-Aware Dynamic Fine-Tuning for Large Language Models

    cs.CL 2026-04 conditional novelty 4.0

    CADFT improves supervised fine-tuning of large language models by dynamically down-weighting training samples whose low model-likelihood indicates high gradient variance, yielding better stability and generalization.