Fine-Tuning

Follow this topic Ask about this topic

Latest papers 205

31 papers in the last four weeks, up 72% on the four weeks before. 0.2% of all new papers.

Jun 4, 2026cs.LG

High-Dimensional Theory of LoRA Fine-Tuning in a Solvable Attention Model

We develop a high-dimensional statistical theory of low-rank adaptation (LoRA) in attention models, capturing the interplay between pre-training and fine-tuning. We introduce a solvable framework in which a single-head attention layer is first pre-trained on a data-abundant task and subsequently adapted via a rank-one LoRA update on limited data. In the high-dimensional limit, both stages admit a sharp asymptotic characterization in terms of a finite set of order parameters, yielding explicit predictions for test errors and representation alignment. Our analysis shows that the impact of pre-training on LoRA is summarized by an effective noise term, from which we derive prescriptions for the optimal pre-training procedure. We also demonstrate a regime with a mismatch between the value of the test error and representation quality, and propose an application of our theory to active fine-tuning.

O. Duranthon, F. Boncoraglio, L. Zdeborová

Statistical Physics of Computation Laboratory, École Polytechnique Fédérale de Lausanne (EPFL) CH-1015 Lausanne

Fine-Tuning Low-Rank Adaptation Attention Mechanisms

Jun 3, 2026cs.LG

Dominant-Layer ZO: A Single Layer Dominates Zeroth-Order Fine-Tuning of LLMs

Zeroth-order (ZO) optimization enables memory-efficient fine-tuning of large language models (LLMs) using only forward passes, but it remains unclear how useful adaptation is distributed across layers. In this work, we reveal a surprising phenomenon: ZO fine-tuning is sharply dominated by a single decoding layer. Across multiple LLM families and downstream tasks, fine-tuning this dominant layer alone consistently matches or even exceeds full-model ZO fine-tuning. We further show that the dominant layer is task-agnostic but model-specific, and can be identified before training through a simple inference-only analysis of activation outliers. Specifically, the dominant layer consistently aligns with the first activation-outlier layer in the pre-trained model. To explain this phenomenon, we analyze how perturbation effects propagate under ZO optimization. We find that the dominant layer combines two key properties: high perturbation sensitivity and early placement in the residual stream, allowing perturbation-induced effects to propagate and accumulate through remaining subsequent decoding layers. As a result, this layer produces disproportionately strong and stable optimization signals under forward-only updates. Extensive experiments on LLaMA2-7B and Qwen3-8B across nine benchmarks show that dominant-layer ZO fine-tuning improves average performance over full-model MeZO and LoRA-based ZO fine-tuning while achieving up to 4.52

\times

training speedup.

Wanhao Yu, Ziyan Wang, Zheng Wang +7

University of North Carolina at Charlotte · University of Houston · State University of New York at Binghamton +2

Fine-Tuning Memory-Efficient Fine-Tuning Zeroth-Order Optimization LLM Fine-Tuning Parameter-Efficient Fine-Tuning

Jun 3, 2026cs.LG

Towards Pretraining Text Encoders for TabPFN

Tabular foundation models, such as TabPFN, achieve strong performance on tabular datasets with numerical and categorical data, but do not natively handle high-cardinality text features. Standard pipelines, therefore, embed text with a language model and compress the resulting vectors with PCA into a small number of scalar features before inputting them into TabPFN. This creates an information bottleneck: most embedding dimensions are discarded, and the compressed representation must then be expanded again by TabPFN's feature encoder. End-to-end alternatives can avoid PCA, but they require large amounts of pretraining data containing text cells and usually perform subpar compared to tabular foundation models that were pretrained on large amounts of synthetic data. Inspired by modality-alignment approaches like LLaVA (vision-to-LLM token projection) and TableGPT-style systems (table-to-LLM token projection), we introduce the TabPFN Text Adapter (text-to-TFM token projection). We freeze both the sentence encoder and TabPFN, and train only a lightweight adapter that maps text embeddings into a short sequence of tokens in TabPFN's embedding space. This design removes the PCA bottleneck, preserves TabPFN's numerical strengths, and is more efficient to train than end-to-end text-tabular pipelines.

Mustafa Tajjar, Alexander Pfefferle, Lennart Purucker +1

Department of Computer Science, University of Freiburg, Freiburg, Germany · ELLIS Institute Tübingen, Tübingen, Germany · Prior Labs, Freiburg, Germany.

Tabular Foundation Models Fine-Tuning Adapter Tuning Tabular ML Tabular Prediction

Jun 3, 2026cs.LG

(Mis)generalization of Helpful-only Fine-tuning

Helpful-only models, that is, models that are trained to always follow user intent, are valuable for dangerous capability evaluations and other areas of AI R&D where refusals would be an obstacle. Little is known about the generalization properties of helpful-only training: helpful-only models refuse less than their harmless counterparts, but previous work has not studied other dimensions of their alignment. We study the shortcomings of existing helpful-only models. We find that some show emergent misalignment, others have residual refusal behaviors, and most show poor steerability, sycophancy, and incoherent character. We show that simple anti-refusal training can cause many of these issues. None of these problems are necessary consequences of helpful-only training, though: we show that synthetic document fine-tuning and adding character-related questions to SFT and RL can mitigate them.

Mohammad Omar Khursheed, Baram Sosis, Fabien Roger

Anthropic Fellows Program MATS · Anthropic

Supervised Fine-Tuning LLM Sycophancy Fine-Tuning LLM Alignment LLM Refusal Behavior

Jun 2, 2026cs.CL

AutoTail-BSFGM: Class-Balance-Aware Fine-Tuning for Chinese Scholarly Text Classification

Scholarly text classification supports literature organization, subject indexing, and research intelligence, but Chinese scholarly corpora often contain imbalanced and semantically adjacent disciplinary labels. We propose AutoTail-BSFGM, a class-balance-aware fine-tuning method that combines an automatically gated tail-prior adjustment, a weak Balanced Softmax auxiliary loss, and Fast Gradient Method adversarial regularization. The method changes only the training objective and procedure; inference uses the same single base-size encoder and linear classifier as the corresponding label-smoothed baseline. We evaluate the method on two CSL-based tasks: an abstract-to-discipline task with 67 labels and a title-to-category task with 13 categories. On the primary abstract task, AutoTail-BSFGM improves validation and lockbox accuracy under both Chinese RoBERTa-WWM and MacBERT-base. With MacBERT-base, validation accuracy increases by 0.83 percentage points and lockbox accuracy by 0.49 points, with a pooled paired McNemar signal on validation (p = 0.023). On the title task, the method improves validation accuracy by 0.70 points and validation balanced accuracy by 2.64 points; lockbox accuracy is approximately neutral while lockbox balanced accuracy improves by 1.22 points. The results support a bounded contribution: AutoTail-BSFGM improves class-balance-sensitive behavior and yields consistent gains for abstract-based scholarly classification, without uniformly improving every metric on every split.

Anling Xiang, Yuwen Yang, Yang Shen

Department of Intelligent Communication, School of Journalism and Communication, Minzu University of China, Beijing, China · ZeeLin (Beijing) Technology Co., Ltd., Beijing, China · School of Journalism and Communication, Tsinghua University, Beijing, China +1

Class-Imbalanced Learning Long-Tailed Classification Fine-Tuning Text Classification Classification

Jun 2, 2026cs.LG

DECA: Decentralizing Block-Wise Adam for Efficient LLM Full-Parameter Fine-Tuning on Non-IID Data

Fine-tuning large language models (LLMs) in privacy-sensitive and resource-constrained environments remains challenging. Since training data are often distributed across multiple clients, decentralized fine-tuning offers a natural paradigm for collaborative adaptation without a central server. However, enabling full-parameter fine-tuning (FPFT) in this decentralized setting is difficult: FPFT provides strong adaptation capacity but incurs prohibitive resource consumption for billion-scale models. Existing decentralized LLM fine-tuning methods therefore mainly rely on parameter-efficient updates, which improve efficiency but may restrict downstream performance. Moreover, client data are typically non-IID, making decentralized optimization more vulnerable to client drift and unstable convergence. To address these challenges, we propose DECA, a resource-efficient decentralized FPFT framework for LLMs on non-IID data. DECA partitions model parameters into disjoint blocks and performs sequential block-wise Adam optimization, reducing resource consumption while preserving decentralized full-parameter adaptation. To stabilize training, DECA further introduces first- and second-order block-wise moment estimates with fresh local gradient statistics and consensus-derived discrepancy signals. We provide rigorous theoretical analysis and extensive experiments, showing that DECA achieves fast convergence, strong downstream performance, and significant resource efficiency.

Yunsheng Yuan, Shaowei Li, Kai Wang +5

School of Computer Science and Technology, Shandong University, Qingdao China · School of Mathematical Science, Peking University, China · IEIT SYSTEM, China +2

Fine-Tuning LLM Fine-Tuning Decentralized Learning

Jun 1, 2026cs.LG

GRZO: Group-Relative Zeroth-Order Optimization for Large Language Model Fine-Tuning

Zeroth-order (ZO) optimization is a memory-efficient alternative to backpropagation for fine-tuning large language models, but its deployment is limited by the high variance of gradient estimation. We propose GRZO, a Group-Relative Zeroth-Order optimizer that draws one pseudo-independent perturbation per mini-batch example and aggregates the per-example losses through group-relative normalization, raising the effective gradient-direction count from one to the batch size at no additional forward cost while preserving inference-level memory. We prove that GRZO is directionally unbiased with variance shrinking proportionally to the batch size, yielding a tighter nonconvex convergence bound than MeZO. Across RoBERTa-large, Llama3-8B, and OPT-13B over multiple tasks, GRZO improves average accuracy on Llama3-8B by

+3.0

over MeZO at

23\%

lower peak GPU memory; as a drop-in replacement for the MeZO core, it lifts sparse, low-rank, and quantized ZO variants by

+6.0

on average.

Liyan Tan, Yequan Zhao, Yifan Yang +3

University of California, Santa Barbara

Fine-Tuning Memory-Efficient Fine-Tuning Zeroth-Order Optimization LLM Fine-Tuning Memory-Efficient Optimization Memory-Efficient Training

Jun 1, 2026cs.CV

Quantifying and Mitigating Domain Shift in Peach Leaf Damage Classification: Attention Mechanisms and Fine-Tuning Strategies

Deep learning models for crop damage assessment are typically trained and validated on curated public imagery, yet their behaviour when deployed in real orchards remains poorly quantified. This work measures and mitigates that gap for peach leaf damage classification, where climate-driven abiotic and biotic stresses produce visually similar foliar symptoms. A benchmark of 1366 manually annotated peach leaves covering six damage types was assembled from public sources, and a second, independently acquired dataset of 180 field images across four classes was collected in a commercial orchard as an unseen target domain. Eleven convolutional backbones and three attention-enhanced variants were compared; CBAM-EfficientNetB5 achieved the best source-domain performance (93.3% accuracy, 0.849 macro F1). Applied directly to the target domain, source-trained models lost on average 0.21 macro F1 points (26.5% relative), with 12 of 14 architectures degrading, confirming that benchmark performance substantially overestimates field behaviour. Three fine-tuning strategies were then evaluated as mitigation: feature extraction proved insufficient in nearly all cases, whereas full fine-tuning recovered performance, with CBAM-EfficientNetB3 reaching 0.9459 accuracy and 0.9297 macro F1 on the local domain. Attention mechanisms improved minority-class recall and adaptation efficiency, but did not by themselves confer robustness to domain shift. The results establish a transferability baseline for peach leaf diagnosis and quantify the adaptation cost of moving from public benchmarks to operational orchards.

Adrián Cánovas-Rodriguez, Miguel A. González-Illán, Maria Fernanda García-Cruz +5

Department of Information and Communication Engineering, University of Murcia, Murcia, Spain · Department of Irrigation, Centro de Edafología y Biología Aplicada del Segura CEBAS-CSIC, Murcia, Spain

Agricultural Image Analysis Fine-Tuning Distribution Shift Cross-Domain Transfer Learning Attention Mechanisms

Jun 1, 2026cs.CV

Parameter-Efficient Fine-Tuning of Large Pretrained Models for Instance Segmentation Tasks

Research and applications in artificial intelligence have recently shifted with the rise of large pretrained models, which deliver state-of-the-art results across numerous tasks. However, the substantial increase in parameters introduces a need for parameter-efficient training strategies. Despite significant advancements, limited research has explored parameter-efficient fine-tuning (PEFT) methods in the context of transformer-based models for instance segmentation. Addressing this gap, this study investigates the effectiveness of PEFT methods, specifically adapters and Low-Rank Adaptation (LoRA), applied to two models across four benchmark datasets. Integrating sequentially arranged adapter modules and applying LoRA to deformable attention--explored here for the first time--achieves competitive performance while fine-tuning only about 1-6% of model parameters, a marked improvement over the 40-55% required in traditional fine-tuning. Key findings indicate that using 2-3 adapters per transformer block offers an optimal balance of performance and efficiency. Furthermore, LoRA, exhibits strong parameter efficiency when applied to deformable attention, and in certain cases surpasses adapter configurations. These results show that the impact of PEFT techniques varies based on dataset complexity and model architecture, underscoring the importance of context-specific tuning. Overall, this work demonstrates the potential of PEFT to enable scalable, customizable, and computationally efficient transfer learning for instance segmentation tasks.

Nermeen Abou Baker, David Rohrschneider, Uwe Handmann

Ruhr West University of applied sciences · Lützowstraße 5, 46236 Bottrop, Germany: Ruhr West University of Applied Sciences

Fine-Tuning Instance Segmentation Adapter Tuning Low-Rank Adaptation Deformable Attention Parameter-Efficient Fine-Tuning

Jun 1, 2026cs.CV

Attention-guided Fine-tuning of Multimodal Large Language Models Improves Chain-of-Thought Reasoning

The effectiveness of Chain-of-Thought (CoT) prompting in Multimodal Large Language Models (MLLMs) remains uncertain: across several visual reasoning benchmarks, CoT prompting often degrades performance compared to direct prompting. In this paper, we provide a systematic analysis of CoT behavior in three modern MLLM families across model scales on datasets requiring step-wise visual evidence. Our analysis identifies two recurring failure modes: premature answer commitment and limited direct visual-token access during rationale generation. We further find that standard CoT-style Supervised Fine-Tuning (CoT-SFT) can mitigate these issues only partially, while often increasing reliance on textual priors and reducing counterfactual visual dependence. Motivated by these findings, we propose Attentive-CoT (Att-CoT), an attention-guided fine-tuning objective that encourages CoT trajectories to delay answer commitment while maintaining sustained visual-token access. Att-CoT can be plugged into any CoT-SFT training run without architectural changes. Experiments on three visual reasoning benchmarks across six MLLMs show that Att-CoT enhances CoT performance over standard fine-tuning.

Sanchit Sinha, Guangzhi Xiong, Bohan Liu +2

University of Virginia

Fine-Tuning Multimodal Large Language Models VLM Reasoning Multimodal CoT Reasoning Visually Grounded Reasoning Multimodal Understanding

May 29, 2026cs.LG

CRMA: A Spectrally-Bounded Backbone for Modular Continual Fine-Tuning of LLMs

Sequential fine-tuning of large language models forces a choice: let the shared substrate keep learning and accept catastrophic forgetting, or freeze it after task one and foreclose cross-task refinement. Per-task adapter methods (LoRAHub, AdapterFusion, PackNet, Progressive Networks) take the second path. We introduce CRMA (Constrained Residual Mixing Adapter), a residual adapter whose internal mixing matrix M is doubly-stochastic at every forward pass via Sinkhorn normalization, so by Birkhoff's theorem ||M||_2 <= 1 holds by construction -- a structural bound, not a penalty. CRMA's spectrally bounded backbone provides a continuously trained shared substrate that earlier modular methods could not, while preserving their forgetting guarantees. On Mistral-7B across 5 sequential domains and 3 seeds, modular per-task LoRA on a CRMA backbone reduces loss-relative drift from +42.96% +/- 5.5 (naive sequential fine-tuning) to -0.17% +/- 0.17, with disjoint per-seed ranges, and improves prior-task holdout loss by 1.99% +/- 0.54 over a matched frozen-substrate baseline. Three independent experimental setups (Mistral-7B 4-domain controlled ablation, TinyLlama 3-domain contamination-controlled replication, Mistral-7B cross-domain probes at 7B) all show positive backward transfer -- without replay buffers, without growing per-task memory, and without distillation. An inference-time ablation on Gemma-2-9B confirms CRMA mediates access to sequentially trained knowledge: 98/100 vs. 38/100 on the same weights and same questions with only CRMA injection toggled. 867 logged training steps verify ||M||_2 = 1.0 within float32 precision (max deviation 1.2 x 10^-7). The forgetting-prevention effect holds across 1.1B-9.2B parameters and four architecture families.

Kiran Nayudu, Aswini Nutakki, Sai Vinay Naidu +1

ModelBrew AI

Continual Learning for LLMs Fine-Tuning Adapter Tuning Parameter-Efficient Fine-Tuning Catastrophic Forgetting

May 29, 2026cs.CL

Subtitle-Aligned Fine-Tuning of Whisper for Swiss German ASR: Benchmark Contamination, Convention Mismatch, and an Honest Baseline at 25.6% WER (13.8% cWER)

We present a systematic study of fine-tuning OpenAI's Whisper large-v3 for Swiss German ASR, using 1,367 hours of broadcast speech paired with Standard German subtitles as weak supervision. Through 16 iterative training runs on an NVIDIA DGX Spark (Grace Blackwell, 128 GB unified memory, up to 1 PFLOP FP4), we compare LoRA and full fine-tuning of the 1.55B-parameter model, investigate hallucination root causes, and quantify the effect of data quality, subtitle alignment, and training strategy. Our best model achieves 25.6% measured WER on the All Swiss German Dialects Test Set (ASGDTS) in an honest evaluation on strictly disjoint data. A harmonized error analysis separating genuine errors from valid stylistic variation (tense, word order, Swiss orthography) yields a content WER (cWER) of 13.8%, counting only actual recognition failures. Bias-corrected estimation reduces this to 8.5%, suggesting the true error rate is roughly one third of measured WER. We demonstrate that published state-of-the-art Swiss German ASR results (17.1-17.5% WER) are inflated by benchmark contamination: a vanilla Whisper model self-trained on the ASGDTS test set with zero Swiss German data achieves 13.88% WER, surpassing all published systems. Experiments with Phi-4-multimodal show an even stronger memorization effect (3.9% WER), revealing that the benchmark primarily measures convention matching rather than dialectal comprehension. We release two models, a LoRA adapter (25.32% WER, 13.9% cWER) and a full fine-tuned model (25.60% WER, 13.8% cWER), among the few publicly available, honestly evaluated Whisper models for Swiss German, under Apache 2.0 with full reproducibility, requiring no institutional data agreements.

Felix Akeret

Independent Researcher, Zurich, Switzerland · Dipl. Ing. ETH Zürich, Dipl. NDS BWI ETH Zürich. · Dipl. NDS BWI ETH Zürich.

ASR Evaluation Fine-Tuning Automatic Speech Recognition Data Contamination in Language Models

May 29, 2026cs.LG

RAFT: Data Refinement and Adaptive Distillation for Domain Fine-Tuning with Alleviated Forgetting

Domain-specific supervised fine-tuning (SFT) often improves in-domain performance at the cost of degrading a model's general capabilities. We view this degradation through two practical gaps in domain SFT: a supervision-compatibility gap, where domain targets differ in style and reasoning format from the original model's natural responses, and a trajectory-preservation gap, where teacher-forced SFT optimizes fixed target tokens without constraining the model's behavior on its own generated prefixes. This process fails to preserve the model's original behavior. We propose RAFT (Data Refinement and Adaptive Distillation for Domain Fine-Tuning with Alleviated Forgetting), a two-stage framework that addresses both factors. First, RAFT constructs model-compatible supervision through self-conditioned rewriting, semantic filtering, and answer fusion. Second, RAFT performs Answer-Conditioned On-Policy Distillation, where the original instruction-tuned model provides soft targets on student-generated trajectories while being conditioned on the fused answer as helpful context. We further introduce top-K temperature distillation and EMA-based adaptive loss balancing to stabilize the domain-general trade-off. Across three instruction-tuned backbones and five domains, RAFT improves average domain accuracy by 23.2% over standard SFT, while recovering part of the SFT-induced degradation on MS-Bench and IFEval, with relative improvements of 18.2% and 10.2%, respectively. These results show that coupling data refinement with trajectory-level preservation provides an effective recipe for domain fine-tuning with alleviated forgetting.

Yuduo Li, Xiaofeng Shi, Qian Kou +2

Beijing Academy of Artificial Intelligence (BAAI) · Beijing Jiaotong University (BJTU)

Supervised Fine-Tuning Fine-Tuning Domain Adaptation for LLMs On-Policy Distillation Catastrophic Forgetting

May 28, 2026cs.LG

Efficient Test-Time Finetuning of LLMs via Convex Reconstruction and Gradient Caching

Test-time finetuning (TTFT) is a rapidly evolving paradigm that adapts a language model to each prompt by retrieving related sequences, updating the model on them, and then evaluating the prompt. However, TTFT is only practical if it is fast: selection and finetuning both happen per query, making each a direct bottleneck. Existing methods trade speed for quality: fast retrieval is often redundant, while stronger diversity-aware selection adds prohibitive per-query cost. We introduce HullFT, a geometric approach to TTFT that addresses both bottlenecks. Given a query, HullFT first represents the query embedding as a sparse convex combination of few training sequences, using efficient projection-free Frank-Wolfe optimization. This yields a support set that is inherently relevant and diverse. We then convert the fractional convex weights into an exact integer multiset for finetuning through a geometric integerization procedure. The resulting multiplicities naturally create repeated examples, which we exploit with Gradient Reuse to amortize forward-backward computation across repeated finetuning steps. Our experiments show that HullFT improves the quality-efficiency tradeoff over current state-of-the-art TTFT methods, achieving lower bits-per-byte at substantially lower total runtime.

Alaa Khamis, Alaa Maalouf

Department of Computer Science University of Haifa

LLM Inference Efficiency Fine-Tuning Test-Time Optimization LLM Fine-Tuning

May 28, 2026cs.CV

A Mechanistic Analysis of Adversarial Fine-tuning of Vision Transformers

The widespread use of image classification models in high-risk, real-world situations necessitates making these models robust to slight disturbances or perturbations, such as blurring or sharpening, in the input images. While vision transformers (ViTs) play an integral role in many modern-day multi-modal models like Vision-Language-Models (VLMs) and Vision-Language-Action (VLA) models, they have received a lack of attention in the setting of robustness. In this work, we analyze the effects of adversarial fine-tuning, a popular method for improving model robustness to image perturbations, on a ViT's performance on perturbed and regular images through a mechanistic lens. We adversarially train a ViT on low-frequency and high-frequency image corruptions, and attempt to explain changes in downstream model performance through an examination of the model's attention mechanisms, internal representations, and knowledge evolution. Overall, our results suggest that, while fine-tuning on inputs with common corruptions improves model performance and certainty on new instances of corrupted data, these improvements do not transfer to other classes of corruptions not seen in the training. Additionally, despite observing changes in visual attention and knowledge evolution across layers, we found that adversarial training did not lead to fundamental changes in the sparse representations learned by ViTs.

Hannah Gao, Isha Agarwal, Dylan Hadfield-Menell +1

Massachusetts Institute of Technology · MIT Cambridge, MA

Adversarial Training Fine-Tuning Image Corruption Robustness Vision Transformer Adversarial Robustness

May 28, 2026cs.AI

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

Supervised fine-tuning (SFT) followed by reinforcement learning (RL) has become a standard post-training paradigm for large language models. This paradigm provides a cold-start for RL exploration, avoiding the inefficiency of pure RL where on-policy sampling yields insufficient positive samples. However, in practice, existing approaches often use a small amount of data for SFT initialization compared to the RL phase, which can cause the model to fit the limited samples and shift away from its pre-trained distribution. This distribution shift impedes the model's ability to effectively explore during subsequent RL training. To address this challenge, we propose that in low-data regimes, SFT should prioritize activating task-relevant capabilities rather than memorizing specific content. Along this line, we propose EKSFT (Entropy-KL Selective Fine-Tuning), which selectively masks tokens that exhibit either high entropy or high KL divergence from a reference model. By excluding these high-uncertainty, distribution-shifting tokens from imitation, EKSFT injects task-specific knowledge while preserving the integrity of the model's pre-trained distribution. Empirical evaluations on mathematical reasoning benchmarks demonstrate that EKSFT consistently outperforms standard SFT. Further RL fine-tuning from the EKSFT model yields consistently better post-RL performance, indicating improved exploration for the RL stage. Our codes and datasets are available at https://github.com/MINE-USTC/EKSFT.

Qi Liu, Mingdi Sun, Yongyi He +5

University of Science and Technology of China · Huawei Cloud

Supervised Fine-Tuning Fine-Tuning LLM Fine-Tuning Selective Fine-Tuning

May 28, 2026cs.AI

Diagnosing Harmful Continuation in Answer-Correct Long-CoT Training Traces

Long chain-of-thought (CoT) traces are widely used as supervision for reasoning-oriented LLM SFT, yet answer-correct traces can still lead to markedly different fine-tuning outcomes. We study post-conclusion continuation in answer-correct long-CoT data: a continuation where the answer appears sufficiently supported, but the trace continues with additional reasoning that remains in the supervised target. To test its training effect, we use a delete-only editor to construct answer-preserving suffix removal and compare CoT-based SFT on the original and processed traces. We observe improved SFT outcomes after removing the editor-identified post-conclusion continuation, suggesting that this continuation is harmful to training in our setting. We therefore refer to this empirically supported phenomenon as harmful continuation. Beyond this intervention, we further characterize the removed post-conclusion continuation through uncertainty and hidden-state progress. We observe persistent local uncertainty together with weakened terminal-directional progress, forming an uncertainty--geometry mismatch. Finally, we instantiate Harmful Continuation Cut (HCC), a lightweight boundary proxy that approximates the editor-identified post-conclusion continuation boundary.

Chen He, Yuhao Wu, Lei Wang +2

University of Electronic Science and Technology of China · Singapore University of Technology and Design · Singapore Management University

Fine-Tuning LLM Fine-Tuning LLM Post-Training LLM Reasoning Reasoning Trace Analysis

May 27, 2026cs.LG

PEFT-Arena: Understanding Parameter-Efficient Finetuning from a Stability-Plasticity Perspective

Parameter-efficient finetuning (PEFT) has become the standard approach for adapting large language models, yet evaluations largely emphasize downstream accuracy while overlooking the retention of pretrained capabilities. We argue that PEFT should be assessed through the stability-plasticity dilemma: the trade-off between target-task adaptation and resistance to forgetting. We introduce PEFT-Arena, a benchmark that jointly measures downstream performance and general capability retention. Across methods, we find distinct stability-plasticity profiles; under comparable parameter budgets, orthogonal finetuning achieves the most favorable Pareto frontier. To explain these differences, we analyze PEFT updates from two geometric perspectives. In weight space, spectral analysis reveals how parameterizations interact with the pretrained singular-value structure. In activation space, retention metrics show whether finetuning preserves or distorts general-capability representations, with forgetting linked to non-isometric representation distortion. Finally, an analysis shows that final SFT checkpoints often overshoot a better target-retention operating point. Inspired by this, we present case studies of a post-hoc improvement with path-wise rewinding.

Yangyi Huang, Ruotian Peng, Zeju Qiu +4

The Chinese University of Hong Kong · Westlake University · MPI for Intelligent Systems

LLM Evaluation Fine-Tuning Parameter-Efficient Fine-Tuning Catastrophic Forgetting

May 26, 2026cs.LG

ReMoE: Boosting Expert Reuse through Router Fine-Tuning in Memory-Constrained MoE LLM Inference

Fine-grained Mixture-of-Experts (MoE) models sparsely activate only a subset of experts per token, reducing activated computation while maintaining high model capacity. However, in memory-constrained inference scenarios, only a small set of experts can be cached. Experts not in the cache must be fetched from slow external storage (e.g., UFS), leading to frequent evictions and substantial I/O overhead. We propose ReMoE, a router fine-tuning framework designed to boost token-wise expert reuse. ReMoE biases the router toward recently selected experts, producing temporally stable routing that better matches cache locality constraints. By increasing short-horizon expert reuse, ReMoE reduces expert fetches from storage without adding inference-time computation. Experiments on DeepSeek and Qwen models show that ReMoE improves expert reuse by 26% while maintaining downstream task performance. Real-system evaluations further confirm these benefits, improving output throughput by 8.4% under vLLM GPU-CPU expert offloading and reducing TPOT by 43.6-49.8% under llama.cpp on Jetson Orin NX, corresponding to a 1.77-1.99

\times

decode speedup across diverse workloads. Checkpoints and usage instructions are available at https://github.com/BUAA-OSCAR/ReMoE.

Xiongwei Zhu, Xiaojian Liao, Tianyang Jiang +3

School of Computer Science and Engineering, Beihang University, Beijing 100191, China · Huawei Technologies Ltd.

LLM Inference Efficiency Expert Offloading Fine-Tuning Mixture of Experts Memory-Efficient Optimization Mixture-of-Experts Inference+1

May 25, 2026cs.CL

Creative Quality Alignment: Expert Tacit Knowledge Transfer via Chain-of-Thought Fine-Tuning

This paper provides an empirical implementation of the creative quality metric proposed in Calibrated Surprise (Zou & Xu, 2026a). The question this paper addresses is: does this mathematical claim hold at the engineering level? To make the answer as general as possible, we deliberately choose the strictest engineering conditions: low data cost and a small base model. Training data comes from approximately 100 expert chain-of-thought (CoT) annotations produced by the BC Protocol (Zou & Xu, 2026b). We also identify a data bias: most publicly available alignment datasets are skewed toward craft-related knowledge, while audience modeling and reality-logic coverage are systematically weak. We use the term Creative Quality Alignment (CQA) to describe this class of engineering methods. We also offer a supporting theoretical observation: in an LLM with a single conditional distribution architecture, calibrating the appreciation side automatically transfers to the generation side via architectural duality. This is the structural reason why ~100 CoT examples are sufficient -- not a purely empirical observation like LIMA (Zhou et al., 2023).

Bo Zou, Chao Xu

Fine-Tuning LLM Alignment Preference Alignment

May 25, 2026cs.CL

Thaka at KSAA-2026 Task 2: Regularized Fine-Tuning for Arabic Speech Diacritization

We describe the winning system for Task 2 of the KSAA-2026 Shared Task on Arabic Speech Dictation with Automatic Diacritization. The task requires producing fully diacritized Arabic text from speech audio and undiacritized transcripts, with only 2,327 training samples available and no external data permitted. Our system fine-tunes CATT-Whisper, a character-level multimodal model combining a pretrained CATT text encoder with a frozen Whisper speech encoder. The key to our approach is training regularization: R-Drop consistency regularization, Optuna-optimized hyperparameters with high weight decay, and Focal Loss. At inference, we average 200 stochastic forward passes across four model checkpoints using Monte Carlo Dropout at the softmax probability level. The system achieves 23.26% WER on the primary leaderboard metric (with case endings, including no-diacritic positions), placing 1st among all participants.

Meshal Alamr, Hassan Alqaeri, Abdullah Aldahlawi

Thaka, Advanced AI and Information Technology Riyadh, Saudi Arabia

Fine-Tuning Arabic NLP Audio Understanding Audio-Language Models Low-Resource Speech Recognition

May 25, 2026cs.CL

Fine-Tuning Over Architectural Complexity: Broad-Coverage PII Detection on PIIBench with DeBERTa

Personally identifiable information (PII) detection systems are frequently trained within narrow source or domain boundaries, limiting coverage when deployed on heterogeneous text. We study model fine-tuning on a corrected multi-source PIIBench preparation spanning 82 retained entity types across ten source datasets. We evaluate three DeBERTa-based approaches: direct token classification fine-tuning, a source-conditioned hierarchical model (SC+H), and a three-phase curriculum extension (SC+H+Curr). Against eight published comparator systems on a reproducible 5,000-record held-out subset (test_5k), direct fine-tuned DeBERTa achieves F1 0.6476, while SC+H and the curriculum variant achieve 0.5899 and 0.2772 respectively; the strongest published comparator reaches only 0.1723. Because validation initially favoured SC+H, we perform a final streamed evaluation on the complete 100,002-record held-out split. Direct fine-tuning remains superior, achieving F1 0.6455 versus 0.5894 for SC+H. Entity-level analysis shows that direct fine tuning wins 54 of 82 fine entity types and all ten coarse groups by support-weighted entity F1, while SC+H retains localised advantages on 28 types. The results indicate that diverse task-specific training data and a simple weighted cross-entropy objective contribute more to broad-coverage PII detection than the tested architectural and curriculum complexity.

Pritesh Jha

Fine-Tuning PII Detection

May 24, 2026cs.LG

Tiny Brains, Giant Impact: Uncovering the Keystone Neurons of LLM with Just a Few Prompts

Large language models (LLMs) display strong comprehensive abilities, yet the internal mechanisms that support these behaviors remain insufficiently understood. In this work, we show that across a wide range of open-weight Transformers, a subset of neurons remains consistently highly activated during inference across tasks of multiple capability dimensions. By probing along the cross-task activation strength, an extremely sparse subset is isolated, whose removal causes a collapse in model behavior, which we term keystone neurons. Our analysis reveals that keystone neurons are a stable and intrinsic neuron subset of the model that is largely established during pretraining. The parameters associated with these neurons are tightly calibrated during the training process, and their precise values are critical for the capabilities of the model. Building on these insights, we propose a supervised fine-tuning approach that updates only keystone neurons, achieving task gains comparable to or even better than full-parameter fine-tuning while better preserving performance in other capability dimensions, despite modifying a much smaller number of parameters.

Xiangtian Ji, Yuxin Chen, Zhengzhou Cai +3

National University of Singapore · Beijing University of Posts and Telecommunications · University of Science and Technology of China

Fine-Tuning LLM Interpretability LLM Fine-Tuning Mechanistic Interpretability Selective Fine-Tuning Parameter-Efficient Fine-Tuning

May 20, 2026cs.LG

ChunkFT: Byte-Streamed Optimization for Memory-Efficient Full Fine-Tuning

This work presents \textsc{ChunkFT}, a memory-efficient fine-tuning framework that reformulates full-parameter fine-tuning around a dynamically activated working set. \textsc{ChunkFT} enables gradient computation for arbitrary sub-tensors without modifying the network architecture, providing an algorithmic foundation for optimizing arbitrary sub-networks while avoiding standard dense gradient computation. We provide a theoretical convergence analysis of \textsc{ChunkFT} in the deterministic setting. Empirically, we apply \textsc{ChunkFT} to fine-tune Llama 3-8B and Llama 3-70B using a single RTX 4090-24GB GPU and 2

\times

H800-80GB GPUs, respectively. Full-parameter fine-tuning of a 7B model with a 1K input length requires only 13.72GB of GPU memory. The results demonstrate the effectiveness of \textsc{ChunkFT} in memory usage, running time, and optimization quality. Moreover, downstream evaluations on language understanding, mathematical reasoning, and MT-Bench show that \textsc{ChunkFT} consistently outperforms existing memory-efficient baselines. Notably, \textsc{ChunkFT} achieves performance comparable to, and in some cases exceeding, full-parameter fine-tuning. Our repository is on https://github.com/misonsky/chunk.

Yongkang Liu, Zijing Wang, Mengjie Zhao +7

Northeastern University, China · Shanghai Jiao Tong University, China · CIS, LMU Munich, Germany +2

Fine-Tuning Memory-Efficient Fine-Tuning LLM Fine-Tuning Memory-Efficient Optimization Memory-Efficient Training

May 19, 2026cs.LG

An exponential mechanism based on quadratic approximations for fine-tuning machine learning models with privacy guarantees

Fine-tuning adapts a pretrained machine learning model to a small, sensitive dataset, but this process risks memorizing individual new data points, making the model vulnerable to adversaries who seek to extract sensitive information. In this work, we develop a randomized algorithm based on the exponential mechanism for fine-tuning while ensuring differential privacy. Our key idea is to construct a simple utility function that combines a local quadratic approximation of the pretrained model with information from the new dataset. The resulting exponential mechanism admits exact sampling from a multivariate normal distribution in closed form. We establish theoretical privacy guarantees, sensitivity bounds, and accuracy estimations for our method. We further introduce a random-projection strategy that makes the approach scalable to high-dimensional models. Numerical experiments on the MNIST benchmark and the MIMIC clinical dataset demonstrate competitive performance against existing differentially private fine-tuning techniques.

Hoang Tran, Jorge Ramirez, Jiayi Wang +3

Computer Science and Mathematics Division, Oak Ridge National Laboratory, Oak Ridge, TN, 37831, USA · Departamento de Matem´aticas, Universidad Nacional de Colombia, Medell´ın, Colombia. · Computational Science and Engineering Division, Oak Ridge National Laboratory, Oak Ridge, TN, 37831, USA +1

Fine-Tuning Privacy-Preserving ML Differential Privacy

May 19, 2026cs.RO

D-CLING: Prior-Preserving Depth-Conditioned Fine-Tuning for Navigation Foundation Models

Navigation Foundation Models (NFMs) trained on large cross-embodied datasets have demonstrated powerful generalizability in various scenarios. Adopting in-domain fine-tuning for an NFM efficiently calibrates the visuomotor policy, promising further improvement even in a novel scenario. However, the fine-tuned models still suffer from poor obstacle avoidance or fail to properly reach the provided goals. Furthermore, model updates using a small subset of data typically erode the pre-trained prior, compromising the pre-training generalization. Consequently, fine-tuning deteriorates the capability of the model for robust and accurate navigation. In this work, we present a novel fine-tuning method that leverages large-scale pre-training while efficiently learning in novel setups, such as environments or camera configurations. In particular, inspired by ControlNet, we fine-tune an NFM by attaching a trainable copy of the pre-trained backbone using zero-initialized residual pathways, thereby learning geometric cues. This design enables the model to efficiently acquire in-domain geometry while preserving pre-trained knowledge across various behaviors. Despite its simplicity, our comprehensive evaluation of real-world navigation suggests that our proposal effectively enables robust long-horizon navigation with minimal collisions and human intervention. Additionally, our offline analysis shows that the proposed method maintains or further improves action prediction capabilities beyond the fine-tuned dataset, providing a key insight into continual learning for general navigation. The project page: https://toyotafrc.github.io/DCLING-Proj/

Shintaro Nakaoka, Takayuki Kanai, Kazuhito Tanaka

The authors are with the Frontier Research Center, Toyota Motor Corporation, Toyota, Aichi, Japan.

Visual Navigation Fine-Tuning Robot Navigation Robot Policy Generalization

May 18, 2026cs.LG

LoRA vs. Full Fine-Tuning: A Theoretical Perspective

Fine-tuning adapts a pre-trained model to downstream tasks using a small amount of labeled data. Low-Rank Adaptation (LoRA) is an efficient fine-tuning method that reduces memory and computation costs while often achieving performance close to full fine-tuning. Despite its widespread use, the theoretical behavior of LoRA is not yet well understood. In this paper, we study LoRA in a simple linear regression setting and compare its excess risk with that of full fine-tuning. Our analysis identifies regimes in which LoRA achieves lower excess risk than full fine-tuning in both overdetermined and underdetermined settings. Specifically, our theory predicts that LoRA can outperform full fine-tuning when the difference between the pretraining and the downstream tasks is effectively low-rank. We further show how the choice of LoRA rank affects generalization performance, explaining why using a very small rank can improve test accuracy in certain settings, even though it limits model expressivity. Finally, we support our theoretical results with experiments on practical tasks, suggesting that the identified tradeoffs and insights extend beyond linear regression.

Ali Zindari, Rotem Mulayoff, Sebastian U. Stich

Universität des Saarlandes · CISPA Helmholtz Center for Information Security

Fine-Tuning Low-Rank Adaptation Statistical Learning Theory

May 18, 2026cs.LG

Distilling Linearized Behavior into Non-Linear Fine-Tuning for Effective Task Arithmetic

Task vector composition has emerged as a promising paradigm for editing pre-trained models, enabling model merging through addition and unlearning through subtraction. Fine-tuning in the tangent space of a pre-trained model (linear fine-tuning) has proven effective, as it produces task vectors that are naturally disentangled and resistant to interference. However, linearized models suffer from limited expressivity during training and incur higher computational costs at inference time, which restrict their practical applicability. In this work, we bridge the gap between linear and standard non-linear fine-tuning. We show that linearity with respect to weight perturbations, a property defined in parameter space, can be enforced through constraints in activation space during training. Concretely, we distill hidden representations from a curvature-regularized linearized teacher into a non-linear student trained via conventional fine-tuning. We find that the resulting model inherits key properties of linearized models for task arithmetic, enabling effective composition of task vectors and achieving strong performance across vision and language benchmarks without incurring any inference-time overhead.

Thomas Sommariva, Francesca Morandi, Simone Calderara +1

ImageLab, Department of Engineering “Enzo Ferrari”, University of Modena and Reggio Emilia, Italy · University of Pisa, Italy

Fine-Tuning Task Arithmetic Model Merging

May 17, 2026cs.LG

DP-SelFT: Differentially Private Selective Fine-Tuning for Large Language Models

Large language models (LLMs) are commonly adapted to downstream tasks through fine-tuning, but fine-tuning data often contains sensitive information that may be leaked by the resulting model. Differential privacy (DP) offers formal protection against such leakage, yet DP fine-tuning of LLMs still suffers from substantial utility degradation due to gradient clipping and noise injection. Existing work improves this trade-off by combining DP with parameter-efficient fine-tuning methods such as LoRA, which constrain the form of updates. In this work, we study a complementary direction: selective fine-tuning, which constrains where updates are applied. We propose DP-SelFT, a framework for differentially private selective fine-tuning of LLMs. DP-SelFT addresses three DP-specific challenges in parameter selection: avoiding repeated privacy cost, improving stability under noisy estimates, and selecting parameters that remain useful under clipped and noisy updates. It first constructs a lightweight DP synthetic dataset and performs selection only on this synthetic data, so the selection stage incurs no additional privacy cost. It then conducts layer-level selection by temporarily training candidate layer subsets on a synthetic training split and evaluating them on a synthetic validation split. Crucially, this temporary training is performed under a perturbation regime matched to downstream DP fine-tuning, with worst-case perturbations of the same scale as DP noise. This favors layer subsets that are not only learnable but also robust to noisy private updates. Experiments on benchmark tasks show that DP-SelFT consistently improves the privacy--utility trade-off over existing DP fine-tuning baselines under the same privacy guarantees.

Haichao Sha, Zihao Wang, Yuncheng Wu +2

Renmin University of China · Nanyang Technological University

Fine-Tuning Privacy-Preserving Language Models LLM Fine-Tuning Privacy-Preserving ML Selective Fine-Tuning

May 16, 2026cs.CL

Closing the Gap at CRAC 2026: Two-Stage Adaptation for LLM-Based Multilingual Coreference Resolution

We present our submission to the LLM track of the 2026 Computational Models of Reference, Anaphora and Coreference (CRAC 2026) shared task. With an average CoNLL F1 score of 74.32 on the official test set, our system ranked first in the LLM track, and third overall. Our system is based on the Gemma-3-27b model, fine-tuned using a two-stage strategy with a multilingual base adapter followed by dataset-specific adapters. We represent mention spans by their headword using an XML-inspired format with local reindexing and annotate documents iteratively. These design choices proved effective across languages, document lengths, and annotation guidelines.

Antoine Bourgois, Olga Seminck, Thierry Poibeau

Lattice (CNRS UMR 8094 & ENS-PSL & Universit´e Sorbonne Nouvelle), Montrouge, France

Multilingual Language Models Coreference Resolution Fine-Tuning Adapter Tuning