Runming Yang

Taiqiang Wu, Chaofan Tao, Jiahao Wang et al.

Kullback-Leiber divergence has been widely used in Knowledge Distillation (KD) to compress Large Language Models (LLMs). Contrary to prior assertions that reverse Kullback-Leibler (RKL) divergence is mode-seeking and thus preferable over the mean-seeking forward Kullback-Leibler (FKL) divergence, this study empirically and theoretically demonstrates that neither mode-seeking nor mean-seeking properties manifest in KD for LLMs. Instead, RKL and FKL are found to share the same optimization objective and both converge after a sufficient number of epochs. However, due to practical constraints, LLMs are seldom trained for such an extensive number of epochs. Meanwhile, we further find that RKL focuses on the tail part of the distributions, while FKL focuses on the head part at the beginning epochs. Consequently, we propose a simple yet effective Adaptive Kullback-Leiber (AKL) divergence method, which adaptively allocates weights to combine FKL and RKL. Metric-based and GPT-4-based evaluations demonstrate that the proposed AKL outperforms the baselines across various tasks and improves the diversity and quality of generated responses. Codes are available at \href{https://github.com/wutaiqiang/LLM_KD_AKL}{github}.

Runming Yang, Taiqiang Wu, Yujiu Yang

Knowledge Distillation (KD), aiming to train a better student model by mimicking the teacher model, plays an important role in model compression. One typical way is to align the output logits. However, we find a common issue named mis-instruction, that the student would be misled when the predictions based on teacher logits do not follow the labels. Meanwhile, there is other useful dark knowledge in the logits such as the class discriminability, which is vital for distillation. In this paper, we propose a simple yet effective Logit Calibration (LoCa) method, which calibrates the logits from the teacher model based on the ground-truth labels. The key insight is to correct the prediction (to address the mis-instruction issue) and maintain useful dark knowledge simultaneously. Our proposed LoCa does not require any additional parameters. Empirical results on image classification and text generation tasks demonstrate that LoCa can effectively improve the performance of baselines.

Tao Liu, Taiqiang Wu, Runming Yang et al.

Supervised fine-tuning (SFT) is a fundamental post-training strategy to align Large Language Models (LLMs) with human intent. However, traditional SFT often ignores the one-to-many nature of language by forcing alignment with a single reference answer, leading to the model overfitting to non-core expressions. Although our empirical analysis suggests that introducing multiple reference answers can mitigate this issue, the prohibitive data and computational costs necessitate a strategic shift: prioritizing the mitigation of single-reference overfitting over the costly pursuit of answer diversity. To achieve this, we reveal the intrinsic connection between token probability and semantic importance: high-probability tokens carry the core logical framework, while low-probability tokens are mostly replaceable expressions. Based on this insight, we propose ProFit, which selectively masks low-probability tokens to prevent surface-level overfitting. Extensive experiments confirm that ProFit consistently outperforms traditional SFT baselines on general reasoning and mathematical benchmarks.

Runming Yang, Taiqiang Wu, Jiahao Wang et al.

Knowledge distillation (KD) has been a predominant method for compressing Large Language Models (LLMs). In this paper, we first revisit KD and Low-Rank Adaption (LoRA) and demonstrate that they follow the same paradigm. Inspired by this observation, we propose a parameter-efficient knowledge distillation method, LLM-NEO, which integrates LoRA into KD to improve the efficiency of knowledge transfer. After that, we summarize some valuable guidelines for the hyperparameters in LLM-NEO. Experimental results on compressing Llama 2 and Llama 3.2 show that LLM-NEO outperforms various baselines. Further analysis demonstrates the robustness of the proposed LLM-NEO on variants of LoRA. The code and trained models are available at [Github](https://github.com/yang3121099/LLM-Neo).

Taiqiang Wu, Runming Yang, Tao Liu et al.

Model merging, typically on Instruct and Thinking models, has shown remarkable performance for efficient reasoning. In this paper, we systematically revisit the simplest merging method that interpolates two weights directly. Particularly, we observe that model interpolation follows a three-stage evolutionary paradigm with distinct behaviors on the reasoning trajectory. These dynamics provide a principled guide for navigating the performance-cost trade-off. Empirical results demonstrate that a strategically interpolated model surprisingly surpasses sophisticated model merging baselines on both efficiency and effectiveness. We further validate our findings with extensive ablation studies on model layers, modules, and decoding strategies. Ultimately, this work demystifies model interpolation and offers a practical framework for crafting models with precisely targeted reasoning capabilities. Code is available at \href{https://github.com/wutaiqiang/MI}{Github}.

He Xiao, Runming Yang, Qingyao Yang et al.

Post-training quantization (PTQ) of large language models (LLMs) to extremely low bit-widths remains challenging due to the fundamental trade-off between computational efficiency and model expressiveness. While existing ultra-low-bit PTQ methods rely on binary approximations or complex compensation mechanisms, they suffer from either limited representational capacity or computational overhead that undermines their efficiency gains. We introduce PTQ to Trit-Planes (PTQTP), the first ternary-weight PTQ framework that decomposes weight matrices into structured ternary {-1, 0, 1} trit-planes using 2x1.58-bit representation. PTQTP achieves multiplication-free inference, identical to 1-bit quantization, while maintaining superior expressiveness through its novel structured decomposition. Our approach provides: (1) a theoretically grounded progressive approximation algorithm ensuring global weight consistency; (2) model-agnostic deployment across diverse modern LLMs without architectural modifications; and (3) uniform ternary operations that eliminate the need for mixed-precision or compensation schemes. Comprehensive experiments across LLaMA3.x and Qwen3 model families (0.6B-70B parameters) demonstrate that PTQTP significantly outperforms existing low-bit PTQ methods, achieving 82.4% mathematical reasoning retention versus 0% for competing approaches. PTQTP approaches and sometimes surpasses 1.58-bit quantization-aware training performance while requiring only single-hour quantization compared to 10-14 GPU days for training-based methods. These results establish PTQTP as a practical solution for efficient LLM deployment in resource-constrained environments. The code will be available at https://github.com/HeXiao-55/PTQTP.

Taiqiang Wu, Runming Yang, Jiayi Li et al.

Large language models (LLMs) consistently benefit from further fine-tuning on various tasks. However, we observe that directly tuning the Instruct (i.e., instruction-tuned) models often leads to marginal improvements and even performance degeneration. Notably, paired Base models, the foundation for these Instruct variants, contain highly similar weight values (i.e., less than 2% on average for Llama 3.1 8B). The Base model tends to be a good learner yet a weak backbone without post-training. Therefore, we propose a novel Shadow-FT framework to tune the Instruct models by leveraging the corresponding Base models. The key insight is to fine-tune the Base model, and then \textit{directly} graft the learned weight updates to the Instruct model. Our proposed Shadow-FT introduces no additional parameters, is easy to implement, and significantly improves performance. We conduct extensive experiments on tuning mainstream LLMs, such as Qwen 3 and Llama 3 series, and evaluate them across 19 benchmarks covering coding, reasoning, and mathematical tasks. Experimental results demonstrate that Shadow-FT consistently outperforms conventional full-parameter and parameter-efficient tuning approaches. Further analyses indicate that Shadow-FT can be applied to multimodal large language models (MLLMs) and combined with direct preference optimization~(DPO). Codes and weights are available at \href{https://github.com/wutaiqiang/Shadow-FT}{Github}.

Taiqiang Wu, Yuxin Cheng, Chenchen Ding et al.

Memristor-based analog compute-in-memory (CIM) architectures provide a promising substrate for the efficient deployment of Large Language Models (LLMs), owing to superior energy efficiency and computational density. However, these architectures suffer from precision issues caused by intrinsic non-idealities of memristors. In this paper, we first conduct a comprehensive investigation into the impact of such typical non-idealities on LLM reasoning. Empirical results indicate that reasoning capability decreases significantly but varies for distinct benchmarks. Subsequently, we systematically appraise three training-free strategies, including thinking mode, in-context learning, and module redundancy. We thus summarize valuable guidelines, i.e., shallow layer redundancy is particularly effective for improving robustness, thinking mode performs better under low noise levels but degrades at higher noise, and in-context learning reduces output length with a slight performance trade-off. Our findings offer new insights into LLM reasoning under non-ideality and practical strategies to improve robustness.

Zhen Li, Yupeng Su, Runming Yang et al.

Large language models have achieved significant advancements in complex mathematical reasoning benchmarks, such as MATH. However, their substantial computational requirements present challenges for practical deployment. Model quantization has emerged as an effective strategy to reduce memory usage and computational costs by employing lower precision and bit-width representations. In this study, we systematically evaluate the impact of quantization on mathematical reasoning tasks. Our results demonstrate that aggressive quantization methods like AWQ and GPTQ introduce up to 32.39% accuracy degradation (average 11.31%) on Llama-3 models, particularly in numerical computation and reasoning planning. To address this, we introduce a multidimensional evaluation framework combining qualitative capability analysis and quantitative error assessment. We further develop targeted recovery strategies, showing that fine-tuning quantized models on only 545 task-specific examples for 3 minutes on 4 GPUs effectively restores reasoning capabilities to near full-precision levels. Additionally, our error assessment pipeline achieves 98.9% accuracy in diagnosing and localizing errors across 3,366 failure cases, providing actionable insights for mitigating quantization-induced degradation.

Taiqiang Wu, Runming Yang, Tao Liu et al.

Post-training, which elicits a pretrained Base model into the corresponding Instruct model, is widely considered to be superficial. In this work, we first reinforce this hypothesis by providing novel quantitative evidence from the weight level that the effective rank (eRank) remains negligibly changed. However, this superficiality also suffers a critical trade-off, improving the exploitation capabilities at the cost of limiting its exploration. To tackle this issue, we propose Timber, a simple yet effective training-free method that enhances the exploration capability of the Instruct model while preserving its exploitation. The key insight is to partially revert Instruct towards the paired Base model by subtle yet targeted refinement of the weight deltas. Extensive experiments on Llama and Qwen series demonstrate that Timber consistently improves vanilla Instruct models, particularly on Pass@k performance. Our findings offer new insights into the post-training stage at the weight level and practical strategies to refine the Instruct model without training.