Supervised Fine-Tuning

Also known as SFT

Follow this topic Ask about this topic

Supervised fine-tuning trains a pretrained model to follow examples of desired behavior, from answering questions to reasoning, using tools, and interpreting images, audio, or other domain-specific inputs. Common approaches curate or generate instruction and agent-trajectory data, distill teacher outputs, and use methods such as LoRA or selective token and layer updates to control training cost and influence. Current work also adjusts what the model learns through token weighting, self-distillation, and constraints that preserve its original output distribution. The central tension is gaining reliable task performance without losing general capabilities or limiting later learning; how SFT should complement reinforcement learning depends on the data, task, and model.

Latest papers

214 papers29 papers in the last four weeks, up 12% on the four weeks before. 0.2% of all new papers.

May 11, 2026

Step Rejection Fine-Tuning: A Practical Distillation Recipe

Rejection Fine-Tuning (RFT) is a standard method for training LLM agents, where unsuccessful trajectories are discarded from the training set. In the context of SWE-bench tasks, this corresponds to filtering out runs where the submitted patch does not pass the tests. However, this approach discards unresolved trajectories, even though they form a large portion of all trajectories for hard tasks and even then may be partially correct. In this work, we propose Step Rejection Fine-Tuning (SRFT) - a practical way to leverage these unresolved trajectories. For this, we employ a critic LLM to assess the correctness of each step in a trajectory. Consequently, during training, we mask the loss for erroneous steps while retaining them in the context window. This way we ensure the model learns to recover from errors without reproducing them. Evaluation on SWE-bench Verified shows that while RFT improves the resolution rate by 2.4% by excluding unresolved trajectories, SRFT improves it by 3.7% by filtering them instead of discarding completely, reaching the total resolution rate of 32.2%.

Igor Slinko, Ilia Zavidnyi, Egor Bogomolov +1

JetBrains Research

Supervised Fine-Tuning LLM Agent Training

May 11, 2026

Annotations Mitigate Post-Training Mode Collapse

Post-training (via supervised fine-tuning) improves instruction-following, but often induces semantic mode collapse by biasing models toward low-entropy fine-tuning data at the expense of the high-entropy pretraining distribution. Crucially, we find this trade-off worsens with scale. To close this semantic diversity gap, we propose annotation-anchored training, a principled method that enables models to adopt the preference-following behaviors of post-training without sacrificing the inherent diversity of pretraining. Our approach is simple: we pretrain on documents paired with semantic annotations, inducing a rich annotation distribution that reflects the full breadth of pretraining data, and we preserve this distribution during post-training. This lets us sample diverse annotations at inference time and use them as anchors to guide generation, effectively transferring pretraining's semantic richness into post-trained models. We find that models trained with annotation-anchored training can attain

6 \times

less diversity collapse than models trained with SFT, and improve with scale.

Jacob Mitchell Springer, Madhu Advani, Lukas Aichberger +7

Carnegie Mellon University · Apple · Johannes Kepler University Linz

Supervised Fine-Tuning Diverse Text Generation Language Model Post-Training Mode Collapse

May 11, 2026

FocuSFT: Bilevel Optimization for Dilution-Aware Long-Context Fine-Tuning

Large language models can now process increasingly long inputs, yet their ability to effectively use information spread across long contexts remains limited. We trace this gap to how attention budget is spent during supervised fine-tuning (SFT) on long sequences: positional biases and attention sinks cause the model to allocate most of its attention to positionally privileged tokens rather than semantically relevant content. This training-time attention dilution (the starvation of content tokens in the attention distribution) weakens the gradient signal, limiting the model's ability to learn robust long-context capabilities. We introduce FocuSFT, a bilevel optimization framework that addresses this problem at training time. An inner loop adapts lightweight fast-weight parameters on the training context to form a parametric memory that concentrates attention on relevant content, and the outer loop performs SFT conditioned on this sharpened representation. Both loops apply bidirectional attention over context tokens while preserving causal masking for responses, reducing the causal asymmetry that gives rise to attention sinks and aligning inner-outer behavior. On BABILong, FocuSFT improves accuracy by up to +14pp across 4K--32K context lengths; on RULER, it raises CWE aggregation from 72.9% to 81.1% at 16K; and on GPQA with agentic tool use, it yields a 24% relative gain in pass@1. Attention analysis shows that FocuSFT reduces attention sink mass by 529

\times

and triples context engagement during training. Code: https://github.com/JarvisPei/FocuSFT

Zehua Pei, Hui-Ling Zhen, Xianzhi Yu +3

The Chinese University of Hong Kong · Huawei Technologies Co., Ltd

Supervised Fine-Tuning Self-Attention LLM Fine-Tuning Attention Sinks

May 11, 2026

Modeling Atomic Conformational Ensembles of Proteins via Test-Time Supervision of Boltz-2 on Cryo-EM Density Maps

Knowledge of a protein's atomic conformational ensemble is critical to determining its function, yet state-of-the-art ensemble prediction models are limited by lack of high-quality conformational data from simulation or experiment. Recent advances in heterogeneous reconstruction for cryo-electron microscopy (cryo-EM) have enabled scientists to visualize ensembles of density maps for larger proteins and complexes not typically accessible through simulation, but building atomic models into these maps remains a challenge. Traditionally, ensemble prediction models are trained via a two-stage process: experimental density maps are converted into atomic structural ensembles through model building, after which these structures are used to train sequence-to-atomic ensemble predictors. In this work, we propose a new principle for fine-tuning pre-trained static structure prediction models such as Boltz-2 directly on raw cryo-EM maps, bypassing the two-stage process. We apply this technique to the problem of atomic model building by fine-tuning Boltz-2 to generate atomic conformations from an input ensemble of cryo-EM maps, achieving superior model building accuracy compared to prior work. Beyond overfitting to individual map ensembles, our method, CryoSampler, also shows preliminary evidence of in-domain generalization after fine-tuning, sampling diverse atomic conformations for an unseen sequences within the same protein family without requiring cryo-EM data. These capabilities indicate that CryoSampler holds the potential to train next-generation atomic ensemble prediction models directly on raw cryo-EM measurements.

Jay Shenoy, Miro Astore, Axel Levy +3

Department of Computer Science Stanford University, Flatiron Institute, & SLAC National Lab Stanford, CA 94305, USA · Center for Computational Biology & Center for Computational Mathematics Flatiron Institute New York, NY 10010, USA · Department of Electrical Engineering Stanford University & SLAC National Lab Stanford, CA 94305, USA +2

Supervised Fine-Tuning Protein Structure Prediction AI-Assisted Scientific Research Cryo-Electron Microscopy

May 10, 2026

Memorize Theorems, Not Instances: Probing SFT Generalization through Mathematical Reasoning

Supervised Fine-Tuning (SFT) is widely used for task-specific adaptation, yet recent work shows it systematically undermines reasoning generalization. We argue the root cause is not memorization itself, but its target: vanilla SFT drives models to exploit and memorize spurious surface correlations in problem-solution pairs, leaving them brittle to superficial input variations. To address this, we propose Theorem-SFT, which reorients supervision toward explicit theorem application by teaching models how rules are invoked rather than what answers look like. Theorem-SFT yields consistent gains across benchmarks and model families: +8.8% on MATH (LLaMA3.2-3B-Instruct) and +20.27% on GeoQA (Qwen2.5-VL-7B-Instruct) without modality-specific re-training. Fine-tuning MLP layers alone matches full-layers performance, implicating feed-forward components as the primary locus of reasoning rules. Our findings reframe the debate: Generalization failures stem not from memorization as a mechanism, but from memorizing the wrong inductive targets.

Ruiying Peng, Mengyu Yang, Jing Lei +3

Tsinghua Shenzhen International Graduate School · Huawei Technologies

Supervised Fine-Tuning Memorization in Language Models OOD Generalization Mathematical Reasoning LLM Reasoning

May 9, 2026

Large Language Models for Sequential Decision-Making: Improving In-Context Learning via Supervised Fine-Tuning

Large language models (LLMs) have shown remarkable in-context learning (ICL) capabilities, yet their potential for sequential decision-making remains underexplored. In this paper, we study the ICL capabilities of LLMs in sequential decision-making settings, including Markov Decision Processes (MDPs), Partially Observable MDPs (POMDPs), and Ambiguous POMDPs (APOMDPs). We fine-tune pretrained LLMs to perform few-shot decision-making directly from offline, oracle-labeled trajectories. Our framework enables flexible imitation of policies through supervised fine-tuning (SFT). Theoretically, we focus on linear MDPs and interpret a fine-tuned attention layer as implicitly estimating optimal Q-functions from in-context data. Building on this interpretation, we derive an end-to-end suboptimality bound for the induced policy that separates the in-context estimation error from the training-length bias. Empirically, across synthetic MDP, POMDP, and APOMDP settings, we find that fine-tuned LLMs achieve substantially smaller optimality gaps than in-context-only and random baselines, with especially large gains in longer-horizon, partially observed, and model-ambiguous environments. Together, these results show that supervised fine-tuning provides an effective route to endowing pretrained LLMs with sequential decision-making capabilities from offline data, which is an important advantage in domains such as healthcare where offline data are abundant.

Minmin Zhang, Sina Aghaei, Soroush Saghafian

Harvard University

Supervised Fine-Tuning Sequential Decision Making Partially Observable RL Imitation Learning In-Context Learning

May 9, 2026

Preserving Foundational Capabilities in Flow-Matching VLAs through Conservative SFT

Unconstrained fine-tuning of flow-matching Vision-Language-Action (VLA) models drives dense parameter overwrites, degrading pre-trained capabilities. We present Conservative Supervised Fine-Tuning (ConSFT), an optimization objective that adapts to target distributions while mitigating catastrophic forgetting, requiring zero prior data or architectural overhead. By dynamically scaling learning signals based on model confidence, ConSFT suppresses excessive gradients from low-confidence samples to prevent disproportionate parameter updates, thereby bounding the intrinsic parameter disruption risk. Inspired by reinforcement learning's trust-region clipping, this formulation establishes a progressive learning dynamic to secure target convergence and prior capability retention, maintaining sparse parameter updates without relying on the parallel reference networks required by explicit regularization. We evaluate ConSFT on the LIBERO and RoboTwin benchmarks across state-of-the-art flow-matching VLAs (

π_0

,

π_{0.5}

, and GR00T-N1.6-3B). The method outperforms vanilla SFT in capability retention by an average absolute margin of over 20%, matching the efficacy of data-heavy Experience Replay in a prior-data-free regime. Real-world robotic deployments confirm that ConSFT precludes spatial overfitting during downstream adaptation, preserving pre-trained physical skills while acquiring sequential target tasks.

Tianyi Zhang, Shaopeng Zhai, Haoran Zhang +2

Shanghai Artificial Intelligence Laboratory

Supervised Fine-Tuning Continual Learning for VLA Models Robotic Manipulation Vision-Language-Action Models Catastrophic Forgetting

May 8, 2026

Rotation-Preserving Supervised Fine-Tuning

Supervised fine-tuning (SFT) improves in-domain performance but can degrade out-of-domain (OOD) generalization. Prior work suggests that this degradation is related to changes in dominant singular subspaces of pretrained weight matrices. However, directly identifying loss-sensitive directions with Hessian or Fisher information is computationally expensive at LLM scale. In this work, we propose preserving projected rotations in pretrained singular subspaces as an efficient proxy for Fisher-sensitive directions, which we call Rotation-Preserving Supervised Fine-Tuning (RPSFT). RPSFT penalizes changes in the projected top-

k

singular-vector block of each pretrained weight matrix, limiting unnecessary rotation while preserving task adaptation. Across model families and sizes trained on math reasoning data, RPSFT improves the in-domain/OOD trade-off over standard SFT and strong SFT baselines, better preserves pretrained representations, and provides stronger initializations for downstream RL fine-tuning. Code is available at https://github.com/jinhangzhan/RPSFT.

Hangzhan Jin, Tianwei Ni, Lu Li +3

1Mila - Quebec AI Institute · 2Polytechnique Montréal · 3Université de Montréal +3

Supervised Fine-Tuning Fine-Tuning LLM Fine-Tuning OOD Generalization

May 8, 2026

On Distinguishing Capability Elicitation from Capability Creation in Post-Training: A Free-Energy Perspective

Debates about large language model post-training often treat supervised fine-tuning (SFT) as imitation and reinforcement learning (RL) as discovery. But this distinction is too coarse. What matters is whether a training procedure increases the probability of behaviors the pretrained model could already produce, or whether it changes what the model can practically reach. We argue that post-training research should distinguish between capability elicitation and capability creation. We make this distinction operational by introducing the notion of accessible support: the set of behaviors that a model can practically produce under finite budgets. Post-training that reweights behaviors within this support is capability elicitation; whereas changing the support itself corresponds to capability creation. We develop this argument through a free-energy view of post-training. SFT and RL can both be seen as reweighting a pretrained reference distribution, only with different external signals. Demonstration signals define low-energy behavior for SFT, and reward signals define low-energy behavior for RL. When the update remains close to the base model, the main effect is local reweighting, not capability creation. Within this framework, the central question is no longer whether post-training is framed as SFT or RL, but whether it reweights behaviors already within reach, or instead expands the model's reachable behavioral space through search, interaction, tool use, or the incorporation of new information.

Yuhao Li, Shengchao Liu

Department of Computer Science and Engineering The Chinese University of Hong Kong

Supervised Fine-Tuning Free Energy Principle LLM Post-Training RL Fine-Tuning

May 8, 2026

Goal-Conditioned Supervised Learning for LLM Fine-Tuning

Large language models often require fine-tuning to better align their behavior with user intent at deployment. Existing approaches are commonly divided into online and offline paradigms. Online methods, such as RL-based alignment, can directly optimize outcome quality but typically rely on external reward models and iterative rollouts, making them costly and difficult to deploy in many cases. Offline methods are more efficient, but prevailing approaches such as supervised fine-tuning (SFT) and direct preference optimization (DPO) remain limited: SFT typically collapses graded feedback into binary supervision, while DPO depends on paired preference data that is often unavailable or expensive to construct. In this paper, we propose goal-conditioned supervised learning (GCSL) as an offline fine-tuning framework for LLMs. Our core idea is to treat feedback signals directly as an explicit goal and train the model, purely through supervised learning, to generate responses that achieve that goal. To better exploit graded feedback, we further introduce a novel goal formulation that defines learning as consistently pursuing outcomes above a target quality threshold, rather than imitating samples from a selected high-quality subset. This design mitigates the bounded-learning effect by learning transferable patterns of meeting or exceeding quality thresholds. We also propose natural-language goal representations to further connect these patterns to the LLM's pretrained knowledge and generalization capabilities. We evaluate our method on three tasks: non-toxic generation, code generation, and LLM for recommendation. Results show that our approach consistently outperforms standard offline fine-tuning baselines while retaining the efficiency, scalability, and simple data requirements of supervised learning.

Shijun Li, Kaiwen Dong, Xiang Gao +1

The University of Texas at Austin · Intuit AI Research

Supervised Fine-Tuning LLM Alignment

May 8, 2026

Dr. Post-Training: A Data Regularization Perspective on LLM Post-Training

Data selection methods address a critical challenge in LLM post-training: effectively leveraging scarce, high-fidelity target data alongside abundant but imperfectly aligned general training data. In this work, we move beyond the data-selection framing and introduce Dr. Post-Training (Data-Regularized Post-Training), a novel framework that reconceptualizes general training data as a data-induced regularizer that prevents overfitting to the scarce target objective, rather than serving as a pool for selection. Specifically, our framework proposes that at each training step, construct a feasible set of model update directions using the general training data, and project the model update direction specified by the scarce target data onto that feasible set. Standard training and existing data selection methods arise as special cases with different choices of the data-induced regularizer, and these methods correspond to different points on a bias--variance spectrum with different regularization strength. Building on this view, we propose a family of methods offering a richer design space and more flexible bias--variance tradeoffs. For practical LLM-scale use, we introduce careful system optimizations that realize these methods with minimal overhead. Extensive experiments across SFT, RLHF, and RLVR show that our methods consistently outperform state-of-the-art data selection baselines, and system benchmarks confirm their efficiency.

Pingbang Hu, Xueshen Liu, Z. Morley Mao +1

University of Illinois Urbana–Champaign · University of Michigan

Supervised Fine-Tuning Training Data Selection RL Post-Training LLM Post-Training Language Model Post-Training

May 7, 2026

Optimizer-Model Consistency: Full Finetuning with the Same Optimizer as Pretraining Forgets Less

Optimizers play an important role in both pretraining and finetuning stages when training large language models (LLMs). In this paper, we present an observation that full finetuning with the same optimizer as in pretraining achieves a better learning-forgetting tradeoff, i.e., forgetting less while achieving the same or better performance on the new task, than other optimizers and, possibly surprisingly, LoRA, during the supervised finetuning (SFT) stage. We term this phenomenon optimizer-model consistency. To better understand it, through controlled experiments and theoretical analysis, we show that: 1) optimizers can shape the models by having regularization effects on the activations, leading to different landscapes around the pretrained checkpoints; 2) in response to this regularization effect, the weight update in SFT should follow some specific structures to lower forgetting of the knowledge learned in pretraining, which can be obtained by using the same optimizer. Moreover, we specifically compare Muon and AdamW when they are employed throughout the pretraining and SFT stages and find that Muon performs worse when finetuned for reasoning tasks. With a synthetic language modeling experiment, we demonstrate that this can come from Muon's strong tendency towards rote memorization, which may hurt pattern acquisition with a small amount of data, as for SFT.

Yuxing Liu, Jianyu Wang, Tong Zhang

1UIUC∗ · 2Apple

Supervised Fine-Tuning LLM Fine-Tuning Catastrophic Forgetting

May 7, 2026

Crafting Reversible SFT Behaviors in Large Language Models

Supervised fine-tuning (SFT) induces new behaviors in large language models, yet imposes no structural constraint on how these behaviors are distributed within the model. Existing behavior interpretation methods, such as circuit attribution approaches, identify sparse subnetworks correlated with SFT-induced behaviors post-hoc. However, such correlations do not imply causal necessity, limiting the ability to selectively control SFT-induced behaviors at inference time. We pursue an alternative by asking: can an SFT-induced behavior be deliberately compressed into a sparse, mechanistically necessary subnetwork, termed a carrier, while remaining controllable at inference time without weight modification? We propose (a) Loss-Constrained Dual Descent (LCDD), which constructs such carriers by jointly optimizing routing masks and model weights under an explicit utility budget, and (b) SFT-Eraser, a soft prompt optimized via activation matching on extracted carrier channels, to reverse the SFT-induced behavior. Across safety, fixed-response, and style behaviors on multiple model families, LCDD yields sparse carriers that preserve target behaviors while enabling strong reversion when triggered by SFT-Eraser. Ablations further establish that the sparse structure is the key precondition for reversal: the same trigger optimization fails on standard SFT models, confirming that structure rather than trigger design is the operative factor. These results provide direct evidence that the learned carriers are causally necessary for the behaviors, pointing to a new direction for systematically localizing and selectively suppressing SFT-induced behaviors in deployed models.

Yuping Lin, Pengfei He, Yue Xing +5

Michigan State University · Hippocratic AI · University of Georgia

Supervised Fine-Tuning Language Model Steering LLM Interpretability Causal Interventions in Language Models

May 7, 2026

One Algorithm, Two Goals: Dual Scoring for Parameter and Data Selection in LLM Fine-Tuning

In Large Language Model (LLM) fine-tuning, parameter and data selection are common strategies for reducing fine-tuning cost, yet they are typically driven by separate scoring mechanisms. When a parameter mask and data subset jointly determine restricted fine-tuning, this separation incurs redundant overhead and makes coordinated selection difficult. We cast parameter and data selection as two bilevel selection problems under a common validation objective and derive a shared local response-surrogate scoring rule. Under first- and second-order validation-improvement approximations, parameter importance and data utility emerge as column-wise and row-wise aggregations of a single gradient interaction matrix, yielding a closed-form row-column correspondence for co-extracting both signals. Building on this structure, we propose DualSFT (Dual-Selection Fine-Tuning), a one-shot dual-scoring algorithm that produces a parameter mask and data subset from shared gradient statistics. On 3B-9B LLMs, single-axis DualSFT variants strengthen target-task performance and stability-plasticity trade-offs within their comparison groups, while full DualSFT yields a more favorable joint-constrained trade-off than sequential hybrid baselines under matched budgets.

Xinrui Chen, Liu Yang, Ou Wu

Hangzhou Institute for Advanced Study, University of Chinese Academy of Sciences · College of Intelligence and Computing, Tianjin University

Supervised Fine-Tuning LLM Fine-Tuning Bilevel Optimization Selective Fine-Tuning Parameter-Efficient Fine-Tuning

May 7, 2026

Navigating by Old Maps: The Pitfalls of Static Mechanistic Localization in LLM Post-Training

The "Locate-then-Update" paradigm has become a predominant approach in the post-training of large language models (LLMs), identifying critical components via mechanistic interpretability for targeted parameter updates. However, this paradigm rests on a fundamental yet unverified assumption: can mechanisms derived from current static parameters reliably guide future dynamic parameter updates? To investigate this, we systematically track the structural evolution of Transformer circuits throughout the supervised fine-tuning (SFT) process, revealing the underlying dynamics of task mechanisms. We introduce three novel metrics-Circuit Distance, Circuit Stability, and Circuit Conflict-to analyze circuit evolution across three dimensions: neural migration, semantic stability, and cross-task interference. Our empirical results reveal that circuits inherently exhibit "Free Evolution" during parameter updates. Consequently, static mechanisms extracted from current states inevitably suffer from temporal latency, making them fundamentally inadequate for guiding future states. Moreover, by deconstructing the "illusion of effectiveness" in existing methods, this work underscores the necessity of "foresight" in mechanistic localization and proposes a predictive framework for future research.

Hang Chen, Jiaying Zhu, Hongyang Chen +3

School of Computer Science and Technology Xi’an Jiaotong University · School of Computer Science and Engineering The Chinese University of Hong Kong · Shaanxi Co., Ltd(Xi’an 710077, China) China Mobile Group +1

Supervised Fine-Tuning Transformer Interpretability LLM Interpretability Mechanistic Interpretability Circuits in Language Models LLM Post-Training+1

May 6, 2026

D-OPSD: On-Policy Self-Distillation for Continuously Tuning Step-Distilled Diffusion Models

The landscape of high-performance image generation models is currently shifting from the inefficient multi-step ones to the efficient few-step counterparts (e.g, Z-Image-Turbo and FLUX.2-klein). However, these models present significant challenges for direct continuous supervised fine-tuning. For example, applying the commonly used fine-tuning technique would compromise their inherent few-step inference capability. To address this, we propose D-OPSD, a novel training paradigm for step-distilled diffusion models that enables on-policy learning during supervised fine-tuning. We first find that the modern diffusion models, where the LLM/VLM serves as the encoder, can inherit its encoder's in-context capabilities. This enables us to formulate the training as an on-policy self-distillation process. Specifically, during training, we make the model act as both the teacher and the student with different contexts, where the student is conditioned only on the text feature, while the teacher is conditioned on the multimodal feature of both the text prompt and the target image. Training minimizes the two predicted distributions over the student's own roll-outs. By optimizing on the model's own trajectory and under its own supervision, D-OPSD enables the model to learn new concepts, styles, etc., without sacrificing the original few-step capacity.

Dengyang Jiang, Xin Jin, Dongyang Liu +9

1The Hong Kong University of Science and Technology · 2Z-Image Team, Alibaba Group · 4The Chinese University of Hong Kong +1

Supervised Fine-Tuning Diffusion Model Distillation On-Policy Self-Distillation Few-Step Diffusion Sampling Diffusion Model Fine-Tuning

May 6, 2026

Stabilizing LLM Supervised Fine-Tuning via Explicit Distributional Control

Post-training large language models (LLMs) often suffers from catastrophic forgetting, where improvements on a target objective degrade previously acquired capabilities. Recent evidence suggests that this phenomenon is primarily driven by excessive distributional drift during optimization. Motivated by this perspective, we propose Anchored Learning, a simple framework that explicitly controls distributional updates during offline fine-tuning via a dynamically evolving moving anchor. Instead of matching a fixed reference distribution, the anchor interpolates between the current model and a frozen reference to construct an intermediate target that the model distills toward, transforming global fine-tuning into a sequence of local trust-region updates in distribution space. Theoretically, we prove this anchor-based update admits a linear KL-divergence upper bound per iteration, ensuring a stable transition between model distributions. Extensive experiments on iGSM, MedCalc, and IFEval show that Anchored Learning consistently lies on the Pareto frontier of gain-stability trade-offs, achieving near-optimal performance improvements while substantially reducing degradation compared to strong baselines. For example, while standard SFT suffers from over 53% performance degradation on iGSM and MedCalc, Anchored Learning slashes this drop to under 5% while maintaining near-optimal gains (e.g., 75.2% on iGSM).

Xinyu Wang, Changzhi Sun, Yuanbin Wu +1

School of Computer Science and Technology, East China Normal University, Shanghai, China · TeleAI, Shanghai, China

Supervised Fine-Tuning Continual Learning for LLMs Catastrophic Forgetting

May 5, 2026

OpenSeeker-v2: Pushing the Limits of Search Agents with Informative and High-Difficulty Trajectories

Deep search capabilities have become an indispensable competency for frontier Large Language Model (LLM) agents, yet their development remains dominated by industrial giants. The typical industry recipe involves a highly resource-intensive pipeline spanning pre-training, continual pre-training (CPT), supervised fine-tuning (SFT), and reinforcement learning (RL). In this report, we show that when fueled with informative and high-difficulty trajectories, a simple SFT approach could be surprisingly powerful for training frontier search agents. By introducing three simple data synthesis modifications: scaling knowledge graph size for richer exploration, expanding the tool set size for broader functionality, and strict low-step filtering, we establish a stronger baseline. Trained on merely 10.6k data points, our OpenSeeker-v2 achieves state-of-the-art performance across 4 benchmarks (30B-sized agents with ReAct paradigm): 46.0% on BrowseComp, 58.1% on BrowseComp-ZH, 34.6% on Humanity's Last Exam, and 78.0% on xbench, surpassing even Tongyi DeepResearch trained with heavy CPT+SFT+RL pipeline, which achieves 43.4%, 46.7%, 32.9%, and 75.0%, respectively. Notably, OpenSeeker-v2 represents the first state-of-the-art search agent within its model scale and paradigm to be developed by a purely academic team using only SFT. We are excited to open-source the OpenSeeker-v2 model weights and share our simple yet effective findings to make frontier search agent research more accessible to the community.

Yuwen Du, Rui Ye, Shuo Tang +4

1Shanghai Jiao Tong University

Supervised Fine-Tuning Tool-Augmented Language Model Agents LLM Agent Trajectory Synthesis LLM Agents LLM Agent Training

May 4, 2026

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

Online reinforcement learning with verifiable rewards (RLVR) turns checkable outcomes into a scalable training signal, but it keeps rollout generation, verifier scoring, and reference-policy evaluations on the optimization path. Static weighted supervised fine-tuning (SFT) on precomputed rollouts seems to remove this bottleneck, yet a weighted likelihood is not specified by rewards alone: its sampler and weights induce the policy being fit. This paper identifies the reference-sampled weighted-SFT objective whose induced policy equals the fixed-reference KL-regularized RLVR optimizer. The optimizer is the standard Boltzmann target policy, obtained by exponentially tilting the reference policy by verifier reward. Matching a weighted-SFT induced policy to this target forces density-ratio weights; in the reference-sampled subclass, this reduces uniquely, up to prompt scaling, to the prompt-normalized Boltzmann weight

\exp(r(x,y)/β)/Z(x)

. BOLT, a Boltzmann-Targeted SFT procedure, is the empirical estimator of this projection. The finite one-shot analysis separates the exact stored-support price

β\log(1/π^*(S_N\mid x))

from partition estimation, effective-sample-size variance, generalization, optimization, and approximation errors. This decomposition explains why extra SFT epochs cannot repair missing reference-policy coverage and exposes the temperature--coverage--variance frontier. When coverage needs adaptive sampling, refreshed Boltzmann projections become KL policy mirror descent; finite inner solves enter as additive drift from the exact mirror step. Single-run Qwen experiments provide projection evidence for the target-matched weight, one-shot saturation, refreshed-sampler gains, and optimization-time savings, within the stated single-run scope.

Yao Shu, Chenxing Wei, Hongbin Lin +2

Supervised Fine-Tuning Reinforcement Learning KL-Regularized RL Policy Optimization Reinforcement Learning with Verifiable Rewards

May 1, 2026

Iterative Finetuning is Mostly Idempotent

If a model has some behavioral tendency, such as sycophancy or misalignment, and it is trained on its own outputs, will the tendency be amplified in the next generation of models? We study this question by training a series of models where each model is finetuned on data generated by its predecessor, and the initial model is seeded with some persona or belief. We test three settings: supervised finetuning (SFT) on instruct models, synthetic document finetuning (SDF) on base models, and direct preference optimization (DPO). In the SFT and SDF settings, traits mostly decay or remain constant so that further finetuning cycles do nothing. In rare cases when amplification occurs, it generally comes at the cost of coherence. In the DPO setting, trait amplification can reliably occur when a model is continually trained with a preference for its own outputs, but vanishes when models are reinitialized at each cycle. Overall, our results suggest that amplification most likely comes from continual post-training, and limiting this stage may be an effective defense. For non-RL finetuning, trait amplification is rare and very sensitive to data quantity, making it significantly less likely to occur accidentally. Finally, the amplification-coherence tradeoff serves as a natural deterrent against trait amplification.

Zephaniah Roe, Jack Sanderson, Dang Nguyen +5

University of Chicago

Supervised Fine-Tuning LLM Post-Training Language Model Self-Improvement RL Fine-Tuning

Apr 30, 2026

Diversity in Large Language Models under Supervised Fine-Tuning

Supervised Fine-Tuning (SFT) is essential for aligning Large Language Models (LLMs) with user intent, yet it is believed to suppress generative diversity. Although this reduction is frequently referenced, formal empirical testing of the phenomenon remains limited. The expressiveness of LLMs by itself was addressed by multiple prior methods. Their varying perspectives suggest that deeper investigation could yield further improvements. In this study, we attribute the decline to two primary drivers: the neglect of low-frequency patterns within fine-tuning datasets and the forgetting of preexisting knowledge. Motivated by our theoretical analysis, we develop Tempered Focal (TOFU) loss, a novel objective that addresses both stated challenges simultaneously. Our extensive evaluation confirms at scale that generation breadth narrows after SFT and strengthens the hypothesis explaining this effect. Across multiple models and benchmarks, we demonstrate that TOFU enhances output diversity while preserving high response quality, offering a principled approach to SFT.

Roman Klypa, Oleksandr Cherednichenko

Univ. Grenoble Alpes, CNRS, Grenoble INP, LJK, 38000 Grenoble, France · Department of Mathematics and Mathematical Statistics, Integrated Science Lab, Umeå University, Sweden.

Supervised Fine-Tuning Catastrophic Forgetting

Apr 28, 2026

How Fast Should a Model Commit to Supervision? Training Reasoning Models on the Tsallis Loss Continuum

SFT-then-RLVR is widely used for post-training reasoning models, but why this specific ordering, and why RLVR-only stalls at cold start, have lacked a unifying theoretical account. We provide that account under a unified loss family

J_Q

using the Tsallis

q

-logarithm.

J_Q

is a single-parameter family that interpolates between RLVR (at

q{=}0

, the exploitation pole) and the log-marginal-likelihood over latent trajectories (at

q{=}1

, the density-estimation pole), under which the standard pipeline corresponds to a stepwise

q{=}1 \to 0

schedule. All members share the same per-example gradient direction, differing only by a per-instance amplification

P_θ^{-q}

that reweights each instance independently of the learning rate. Under gradient flow analysis, we show that the exploitation pole requires

Ω(\frac{1}{p_0})

time to escape cold start but is robust to label noise, while the density-estimation pole escapes in

Θ\big(\log(\frac{1}{p_0})\big)

but memorizes label noise. This separation explains how SFT (

q{=}1

) first moves the model out of the cold-start regime, followed by the more robust RLVR (

q{=}0

), under the SFT-then-RLVR paradigm. We further derive two Monte Carlo estimators that directly optimize fixed-

q

on the

J_Q

continuum, without annotated rationales: Gradient-Amplified RL (GARL) and Posterior-Attenuated Fine-Tuning (PAFT), with shared bias

O\big(\frac{q}{M P_θ^q}\big)

but different variance and stability properties. On FinQA, HotPotQA, and MuSiQue, GARL at sufficiently high

q

substantially mitigates cold-start stalling, escaping cold start where GRPO fails entirely. In warm start, GARL at low

q

dominates FinQA where training is stable; on HotPotQA and MuSiQue, GARL destabilizes and PAFT at

q{=}0.75

remains stable, reaching

47.9

m@16 on HotPotQA (

+13.9

over GRPO).

Chu-Cheng Lin, Eugene Ie

Google

Supervised Fine-Tuning Reinforcement Learning RL for Language Model Reasoning Reinforcement Learning with Verifiable Rewards

Apr 27, 2026

Distilling Self-Consistency into Verbal Confidence: A Pre-Registered Negative Result and Post-Hoc Rescue on Gemma 3 4B

Small instruct-tuned LLMs produce degenerate verbal confidence under minimal elicitation: ceiling rates above 95%, near-chance Type-2 AUROC, and Invalid validity profiles. We test whether confidence-conditioned supervised fine-tuning (CSFT) with self-consistency-derived targets can close the gap between internal information and verbal readout. A pre-registered Phase 0 protocol on Gemma 3 4B-it with a modal filter restricting training to items with correct modal answers produced a negative result: AUROC2 dropped from 0.554 to 0.509 due to label-entropy collapse in the training targets. An exploratory rescue removed the filter, training on all 2,000 calibration items. This produced a binary verbal correctness discriminator with AUROC2 = 0.774 on held-out TriviaQA, compressing a 10-sample self-consistency signal (AUROC2 = 0.999) into a single-pass readout exceeding logit entropy (0.701). The shuffled-target control showed no improvement (0.501). On MMLU, accuracy improved from 54.2% to 77.4% with the shuffled model at baseline (56.1%), supporting a target-dependent interpretation. The result is exploratory, binary rather than continuously calibrated, and observed at a single scale. It identifies two design lessons: confidence training requires label entropy, and correct targets regularise output format.

Jon-Paul Cacioli

Independent Researcher, Melbourne, Australia

Supervised Fine-Tuning Confidence Estimation in Language Models Self-Consistency Decoding Language Model Calibration LLM Reliability Language Model Distillation

Apr 24, 2026

Preserving Long-Tailed Expert Information in Mixture-of-Experts Tuning

Despite MoE models leading many benchmarks, supervised fine-tuning (SFT) for the MoE architectures remains difficult because its router layers are fragile. Methods such as DenseMixer and ESFT mitigate router collapse with dense mixing or auxiliary load-balancing losses, but these introduce noisy gradients that often degrade performance. In preliminary experiments, we systematically pruned experts and observed that while certain super experts are activated far more frequently, discarding less used experts still leads to notable performance degradation. This suggests that even rarely activated experts encode non-trivial knowledge useful for downstream tasks. Motivated by this, we propose an auxiliary-loss-free MoE SFT framework that combines bias-driven sparsification with always-active gated condenser experts. Rather than enforcing balanced activation across all experts, our method encourages task-relevant experts to remain active while pushing long-tailed experts toward inactivity. The condenser experts provide a persistent, learnable pathway that alleviates gradient starvation and facilitates consolidation of information that would otherwise remain fragmented across sparsely activated experts. Analysis further suggest that this design better preserves long-tailed expert information under sparse routing. Experiments on large-scale MoE models demonstrate that our approach outperforms state-of-the-art SFT baselines such as DenseMixer and ESFT, achieving average gain of 2.5%+ on both mathematical reasoning and commonsenseQA benchmarks.

Haoze He, Xingyuan Ding, Xuan Jiang +5

Carnegie Mellon University · MIT · UCSD +1

Supervised Fine-Tuning Fine-Tuning Long-Tail Learning Sparse Mixture-of-Experts Mixture-of-Experts Models

Apr 22, 2026

Parallel-SFT: Improving Zero-Shot Cross-Programming-Language Transfer for Code RL

Modern language models demonstrate impressive coding capabilities in common programming languages (PLs), such as C++ and Python, but their performance in lower-resource PLs is often limited by training data availability. In principle, however, most programming skills are universal across PLs, so the capability acquired in one PL should transfer to others. In this work, we propose the task of zero-shot cross-programming-language transfer for code RL. We find that, for Llama-3.1, RL training for code generation in a source PL fails to improve, and sometimes even degrades, the performance on other target PLs. To address this, we hypothesize that effective RL transfer requires a generalizable SFT initialization before RL. We thus propose Parallel-SFT, an SFT strategy that incorporates "parallel programs" -- functionally equivalent code implemented in multiple PLs -- into the data mixture. We demonstrate that this improves transferability: when we subsequently perform RL on our Parallel-SFT model, we observe better generalization to unseen PLs. Analysis of the model internal representations reveals that Parallel-SFT leads to a more functionality-centric latent space, where equivalent programs across PLs are more tightly clustered, which we hypothesize to contribute to the improved transferability.

Zhaofeng Wu, Shiqi Wang, Boya Peng +5

Meta Superintelligence Labs · MIT

Supervised Fine-Tuning RL for Code Generation Zero-Shot Learning Code Generation Code Language Models

Apr 22, 2026

Compatibility-Aware Dynamic Fine-Tuning for Large Language Models

Supervised Fine-Tuning (SFT) is the predominant paradigm for aligning large language models (LLMs), yet it suffers from optimization instability and limited generalization. Recent work attributes this issue to pathological gradient scaling and proposes Dynamic Fine-Tuning (DFT) to correct it at the token level. However, DFT assumes all demonstrations are equally suitable learning targets, an assumption violated by the strong heterogeneity of large-scale instruction data, where demonstration-policy mismatch induces high-variance updates at the sample level. We introduce Compatibility-Aware Dynamic Fine-Tuning (CADFT), a principled extension of DFT that controls sample-level optimization variance. CADFT derives a dynamic, policy-dependent compatibility signal from model likelihoods to modulate supervised updates, suppressing high-variance gradients from incompatible demonstrations. We further propose a delayed, low-frequency compatibility-guided rewriting strategy to transform persistently incompatible demonstrations into learnable targets. We show that CADFT can be interpreted as a variance-controlled estimator that generalizes token-level stabilization in DFT to the sample level. Extensive experiments demonstrate improved stability, generalization, and cold-start reinforcement learning initialization, while remaining fully supervised and independent of explicit reward modeling.

Yucheng Zhou, Junwei Sheng, Qianning Wang +1

SKL-IOTSC, CIS, University of Macau · Auckland University of Technology

Supervised Fine-Tuning Fine-Tuning LLM Fine-Tuning Instruction Tuning

Apr 20, 2026

When Can LLMs Learn to Reason with Weak Supervision?

Large language models have achieved significant reasoning improvements through reinforcement learning with verifiable rewards (RLVR). Yet as model capabilities grow, constructing high-quality reward signals becomes increasingly difficult, making it essential to understand when RLVR can succeed under weaker forms of supervision. We conduct a systematic empirical study across diverse model families and reasoning domains under three weak supervision settings: scarce data, noisy rewards, and self-supervised proxy rewards. We find that generalization is governed by training reward saturation dynamics: models that generalize exhibit a prolonged pre-saturation phase during which training reward and downstream performance climb together, while models that saturate rapidly memorize rather than learn. We identify reasoning faithfulness, defined as the extent to which intermediate steps logically support the final answer, as the pre-RL property that predicts which regime a model falls into, while output diversity alone is uninformative. Motivated by these findings, we disentangle the contributions of continual pre-training and supervised fine-tuning, finding that SFT on explicit reasoning traces is necessary for generalization under weak supervision, while continual pre-training on domain data amplifies the effect. Applied together to Llama3.2-3B-Base, these interventions enable generalization across all three settings where the base model previously failed.

Salman Rahman, Jingyan Shen, Anna Mordvina +3

University of California, Los Angeles · New York University · Google

Supervised Fine-Tuning RL for Language Model Reasoning Weakly Supervised Learning Reinforcement Learning with Verifiable Rewards

Apr 20, 2026

Generalization Boundaries of Fine-Tuned Small Language Models for Graph Structural Inference

Small language models fine-tuned for graph property estimation have demonstrated strong in-distribution performance, yet their generalization capabilities beyond training conditions remain poorly understood. In this work, we systematically investigate the boundaries of structural inference in fine-tuned small language models along two generalization axes - graph size and graph family distribution - and assess domain-learning capability on real-world graph benchmarks. Using a controlled experimental setup with three instruction-tuned models in the 3-4B parameter class and two graph serialization formats, we evaluate performance on graphs substantially larger than the training range and across held-out random graph families. Our results show that fine-tuned models maintain strong ordinal consistency across structurally distinct graph families and continue to rank graphs by structural properties on inputs substantially larger than those seen during training, with distinct architecture-specific degradation profiles. These findings delineate where fine-tuned small language models generalize reliably, providing empirical grounding for their use in graph-based reasoning tasks.

Michal Podstawski

NASK National Research Institute Warsaw, Poland

Supervised Fine-Tuning Small Language Models

Apr 20, 2026

Post-training is (Massive) Supervised Learning

The prevailing paradigm for training LLMs has evolved to rely on a massive post-training phase consisting of SFT and RL. In this position paper, we argue that this methodology effectively marks a reversion to the

pre-train then fine-tune'' approach of the BERT era, explicitly tailoring models to the desired behaviors and specific benchmarks on which they are evaluated. We begin with a historical overview of LLMs, describing the different phases of the LLM evolution. We argue that the current landscape is remarkably similar to the early days of LLMs, where task performance heavily relied on fitting the models to in-distribution datasets. To empirically demonstrate this, we compare pre-trained models to randomly initialized ones, by fine-tuning both variants on modern reasoning datasets and evaluating them on competitive math and code benchmarks. We show that models post-trained from scratch yield highly non-trivial performance. Our findings suggest that current post-training methodologies function primarily as a distribution-fitting mechanism. We finish by positing that developing generally capable models and systems requires moving beyond extensive post-training for predefined behaviors, shifting instead toward training procedures where models

learn how to learn''.

Michael Hassid, Yossi Adi, Roy Schwartz

1FAIR, Meta AI · 2The Hebrew University of Jerusalem

Supervised Fine-Tuning Language Model Post-Training

Apr 20, 2026

Bridging the Reasoning Gap in Vietnamese with Small Language Models via Test-Time Scaling

The democratization of ubiquitous AI hinges on deploying sophisticated reasoning capabilities on resource-constrained devices. However, Small Language Models (SLMs) often face a "reasoning gap", particularly in non-English languages like Vietnamese, where they struggle to maintain coherent chains of thought. This paper investigates Test-Time Scaling strategies for the Qwen3-1.7B architecture within the context of Vietnamese Elementary Mathematics. We introduce Vi-S1K, a high-fidelity reasoning dataset localized via a Gemini 2.5 Flash-Lite powered pipeline, and Vi-Elementary-Bench, a dual-resource benchmark for rigorous evaluation. Using an LLM-as-a-Judge protocol, we reveal that the base model possesses robust latent knowledge (Accuracy: 4.05/5.00) but suffers from a severe "formatting gap" in communication. Supervised Fine-Tuning (SFT) acts as a critical "reasoning unlocker", yielding a 77% improvement in Explanation Quality and bridging the gap between raw calculation and pedagogical coherence. Furthermore, our analysis of prompting strategies uncovers a significant trade-off: structured frameworks like ReAct impose a "cognitive tax" on the 1.7B parameter capacity, degrading performance relative to pure Chain-of-Thought (CoT) combined with Self-Consistency. These findings establish a deployment hierarchy for SLMs, demonstrating that SFT combined with simplified test-time scaling is superior to complex agentic workflows for edge-based reasoning.

Bui The Trung, Do Minh Duc, Nguyen Van Vinh +1

Computer Science and Engineering VNU Vietnam Japan University Hanoi, Vietnam · Information Science JAIST Ishikawa, Japan · Faculty of Information VNU University of Engineering and Technology Hanoi, Vietnam +1

Supervised Fine-Tuning Mathematical Reasoning Benchmarks Test-Time Scaling Small Language Models