A Layer-wise Analysis of Supervised Fine-Tuning

• URL: http://arxiv.org/abs/2604.11838v1
• Date: Sun, 12 Apr 2026 12:57:12 GMT
• Title: A Layer-wise Analysis of Supervised Fine-Tuning
• Abstract summary: Supervised Fine-Tuning (SFT) incurs the risk of catastrophic forgetting.<n>We investigate this mechanism via a comprehensive analysis utilizing information-theoretic, geometric, and optimization metrics.<n>We propose Mid-Block Efficient Tuning, which selectively updates these critical intermediate layers.
• Score: 8.337291809639352
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: While critical for alignment, Supervised Fine-Tuning (SFT) incurs the risk of catastrophic forgetting, yet the layer-wise emergence of instruction-following capabilities remains elusive. We investigate this mechanism via a comprehensive analysis utilizing information-theoretic, geometric, and optimization metrics across model scales (1B-32B). Our experiments reveal a distinct depth-dependent pattern: middle layers (20\%-80\%) are stable, whereas final layers exhibit high sensitivity. Leveraging this insight, we propose Mid-Block Efficient Tuning, which selectively updates these critical intermediate layers. Empirically, our method outperforms standard LoRA up to 10.2\% on GSM8K (OLMo2-7B) with reduced parameter overhead, demonstrating that effective alignment is architecturally localized rather than distributed. The code is publicly available at https://anonymous.4open.science/r/base_sft.

Related papers

• Deeper is Not Always Better: Mitigating the Alignment Tax via Confident Layer Decoding [92.1521161575198]
  We introduce Confident Decoding, a training-free decoding strategy that dynamically selects the most reliable near-final layer.<n> Experiments across dense and Mixture-of-Experts LLMs demonstrate consistent gains on challenging reasoning benchmarks.
  arXiv  Detail & Related papers  (2026-06-20T07:03:26Z)
• Alternating Gradient Flow Utility: A Unified Metric for Structural Pruning and Dynamic Routing in Deep Networks [52.153950303594684]
  We propose a decoupled kinetic paradigm inspired by Alternating Gradient Flow (AGF)<n>AGF accurately captures the network's structural "kinetic utility"<n>We design a hybrid routing framework that decouples AGF-guided offline structural search from online execution via zero-cost physical priors.
  arXiv  Detail & Related papers  (2026-03-12T18:19:21Z)
• ConsistentRFT: Reducing Visual Hallucinations in Flow-based Reinforcement Fine-Tuning [85.20505958752928]
  Reinforcement Fine-Tuning (RFT) on flow-based models is crucial for preference alignment.<n>RFT often introduce visual hallucinations like over-optimized details and semantic misalignment.<n>This work preliminarily explores why visual hallucinations arise and how to reduce them.
  arXiv  Detail & Related papers  (2026-02-03T11:49:46Z)
• HyLRA: Hybrid Layer Reuse Attention for Efficient Long-Context Inference [11.718567830546538]
  Long-context inference in Large Language Models is bottlenecked by the quadratic computation complexity of attention.<n>We introduce bf HyLRA, a novel framework driven by layer-wise sparsity profiling.<n>We show that HyLRA improves inference throughput by 6%--46% while maintaining comparable performance.
  arXiv  Detail & Related papers  (2026-01-31T15:36:17Z)
• Hierarchical Alignment: Surgical Fine-Tuning via Functional Layer Specialization in Large Language Models [4.935224714809964]
  We introduce Hierarchical Alignment, a novel method that applies targeted DPO to distinct functional blocks of a model's layers.<n>Specifically, aligning the local layers (Local-Align) enhances grammatical fluency.<n> aligning the global layers (Global-Align) improves factual consistency as hypothesized but also proves to be the most effective strategy for enhancing logical coherence.
  arXiv  Detail & Related papers  (2025-10-14T00:58:34Z)
• High-Layer Attention Pruning with Rescaling [14.141903038286362]
  Pruning is a highly effective approach for compressing large language models (LLMs)<n>We propose a novel pruning algorithm that strategically prunes attention heads in the model's higher layers.<n>We conduct comprehensive experiments on a wide range of LLMs, including LLaMA3.1-8B, Mistral-7B-v0.3, Qwen2-7B, and Gemma2-9B.
  arXiv  Detail & Related papers  (2025-07-02T17:15:05Z)
• Spectral Insights into Data-Oblivious Critical Layers in Large Language Models [7.486925126518052]
  We introduce a data-oblivious approach to identify intrinsic critical layers in pre-fine-tuned language models.<n>We show that layers with significant shifts in representation space are also those most affected during fine-tuning.
  arXiv  Detail & Related papers  (2025-05-31T04:21:39Z)
• A Sliding Layer Merging Method for Efficient Depth-Wise Pruning in LLMs [13.000188564679998]
  This paper reveals the Patch-like'' feature relationship between layers in large language models by analyzing the correlation of the outputs of different layers in the reproducing kernel Hilbert space.<n>We propose a sliding layer merging method that dynamically selects and fuses consecutive layers from top to bottom according to a pre-defined similarity threshold.<n>Our method outperforms existing pruning techniques in both zero-shot inference performance and retraining recovery quality after pruning.
  arXiv  Detail & Related papers  (2025-02-26T14:15:24Z)
• Determining Layer-wise Sparsity for Large Language Models Through a Theoretical Perspective [55.90119819642064]
  We address the challenge of determining the layer-wise sparsity rates of large language models (LLMs) through a theoretical perspective.<n>This refers to the cumulative effect of reconstruction errors throughout the sparsification process.<n>We derive a simple yet effective approach to layer-wise sparsity allocation that mitigates this issue.
  arXiv  Detail & Related papers  (2025-02-20T17:51:10Z)
• ZARTS: On Zero-order Optimization for Neural Architecture Search [94.41017048659664]
  Differentiable architecture search (DARTS) has been a popular one-shot paradigm for NAS due to its high efficiency. This work turns to zero-order optimization and proposes a novel NAS scheme, called ZARTS, to search without enforcing the above approximation. In particular, results on 12 benchmarks verify the outstanding robustness of ZARTS, where the performance of DARTS collapses due to its known instability issue.
  arXiv  Detail & Related papers  (2021-10-10T09:35:15Z)
• iDARTS: Differentiable Architecture Search with Stochastic Implicit Gradients [75.41173109807735]
  Differentiable ARchiTecture Search (DARTS) has recently become the mainstream of neural architecture search (NAS) We tackle the hypergradient computation in DARTS based on the implicit function theorem. We show that the architecture optimisation with the proposed method, named iDARTS, is expected to converge to a stationary point.
  arXiv  Detail & Related papers  (2021-06-21T00:44:11Z)
• Layer-adaptive sparsity for the Magnitude-based Pruning [88.37510230946478]
  We propose a novel importance score for global pruning, coined layer-adaptive magnitude-based pruning (LAMP) score. LAMP consistently outperforms popular existing schemes for layerwise sparsity selection.
  arXiv  Detail & Related papers  (2020-10-15T09:14:02Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.