arXiv subjects

Yonggan Fu

Publications and source records attributed to Yonggan Fu.

At least 19 recordsLinked to original sources

arXiv2026

Nemotron-Labs-Diffusion: A Tri-Mode Language Model Unifying Autoregressive, Diffusion, and Self-Speculation Decoding

We introduce Nemotron-Labs-Diffusion, a tri-mode language model (LM) that unifies AR, diffusion, and self-speculation decoding within a single architecture. Trained with a joint AR-diffusion objective, Nemotron-Labs-Diffusion can switch modes to sustain high throughput across deployment settings and concurrency levels. Our study shows that (1) AR and diffusion objectives are complementary: diffusion improves lookahead planning, while AR provides left-to-right linguistic priors. (2) In self-speculation mode, diffusion drafts while AR verifies, outperforming multi-token prediction (MTP) methods in both acceptance rate and real-device efficiency. (3) A speed-of-light analysis further demonstrates diffusion's long-term potential, with up to 76.5% more tokens per forward pass than self-speculation under an optimal sampler. Scaling to 3B, 8B, and 14B parameters, our Nemotron-Labs-Diffusion family, including base, instruct, and vision-language models, consistently outperforms state-of-the-art open-source AR and diffusion LMs in both accuracy and speed. For example, Nemotron-Labs-Diffusion-8B decodes 6x more tokens per forward than Qwen3-8B with comparable accuracy, translating to 4x higher throughput on SPEED-Bench with SGLang on a GB200 GPU.

cs.CL↗

arXiv2026

Nemotron-Labs-Diffusion-Image: Advancing Masked Discrete Diffusion for High-Resolution Image Synthesis

We propose Nemotron-Labs-Diffusion-Image, a state-of-the-art masked discrete diffusion model (MDM) for high-resolution text-to-image synthesis. Compared with prior work on masked image generation, Nemotron-Labs-Diffusion-Image addresses two key challenges. First, unlike continuous diffusion models which progressively refine latent representations across the entire image, standard MDMs lack self-correcting capability because discrete tokens cannot be modified once they are unmasked. Second, although increasing the vocabulary size of discrete image tokenizers improves reconstruction fidelity, it introduces optimization difficulties for generative modeling as the per-token training signal becomes increasingly sparse. To address the first challenge, Nemotron-Labs-Diffusion-Image incorporates a token-editing mechanism that enables the model to dynamically revise already-unmasked tokens during inference, similar to how a sculptor iteratively refines their work. To tackle the second challenge, we propose a Grouped Cross-Entropy (GCE) objective that assigns positive learning signals to tokens neighboring the ground truth in embedding space, thereby alleviating signal sparsity. To further improve training efficiency, we implement a custom fused operator for GCE that significantly reduces VRAM usage in large-vocabulary settings. Experimental results demonstrate that these innovations substantially improve both training efficiency and image fidelity of masked discrete image generators, achieving a score of 0.90 on GenEval, 86.9 on DPG and 10.76 of HPSv3.

cs.CV↗

arXiv2026

PRESTO: Prefix-Aligned Tree Drafting for Diffusion Speculative Decoding

Diffusion Large Language Models (dLLMs) have emerged as a promising alternative to autoregressive (AR) LLMs, generating tokens in parallel. This makes them effective draft models for speculative decoding (SD), producing an entire block of draft tokens in a single forward pass. Yet existing diffusion-based drafting methods rely on linear drafting, even though dLLMs emit multiple candidate tokens across positions, inducing a large combinatorial space of decoding paths. Consequently, they limit acceptance length and decoding efficiency. To exploit this multi-candidate structure, we apply tree-based drafting to diffusion drafters, enabling exploration of diverse candidate paths. However, we find that naive tree drafting is suboptimal: diffusion marginals are prefix-blind, mismatching the prefix-based AR verification and yielding unreliable path ranking. We propose PRESTO, a principled framework that extends tree-based drafting to diffusion drafters while resolving the fundamental mismatch between diffusion draft confidence and prefix-based AR verification through PREfix-aligned Scoring and priority-based Tree search for diffusion speculative decOding. The key principles behind PRESTO are that (1) candidate ranking should align with the prefix-based nature of AR verification, and (2) tree construction should prioritize candidate paths with high verification potential to maximize acceptance length. Extensive experiments show that PRESTO achieves up to an average of $1.5\times$ end-to-end throughput speedup on the state-of-the-art dedicated diffusion drafter SD and an average of $1.12\times$ on self-speculative diffusion LLMs across diverse benchmarks.

cs.AI↗

arXiv2026

$R^2$-dLLM: Accelerating Diffusion Large Language Models via Spatio-Temporal Redundancy Reduction

Diffusion Large Language Models (dLLMs) have emerged as a promising alternative to autoregressive generation by enabling parallel token prediction. However, practical dLLM decoding still suffers from high inference latency, which limits deployment. In this work, we observe that a substantial part of this inefficiency comes from recurring redundancy in the decoding process, including spatial redundancy caused by confidence clusters and positional ambiguity, and temporal redundancy caused by repeatedly remasking predictions that have already stabilized. Motivated by these patterns, we propose $R^{2}$-dLLM, a unified framework for reducing decoding redundancy from both inference and training perspectives. At inference time, we introduce training-free decoding rules that aggregate local confidence and token predictions, and finalize temporally stable tokens to avoid redundant decoding steps. We further propose a redundancy-aware supervised fine-tuning pipeline that aligns the model with efficient decoding trajectories and reduces reliance on manually tuned thresholds. Experiments demonstrate that $R^{2}$-dLLM consistently reduces the number of decoding steps by up to 88\% compared to existing decoding strategies, while maintaining competitive generation quality across different models and tasks. These results validate that decoding redundancy is a central bottleneck in dLLMs, and that explicitly reducing it yields substantial practical efficiency gains. Our code and models are available at https://github.com/GATECH-EIC/R2-dLLM.

cs.CL↗

arXiv2026

Fast-dVLM: Efficient Block-Diffusion VLM via Direct Conversion from Autoregressive VLM

Vision-language models (VLMs) predominantly rely on autoregressive decoding, which generates tokens one at a time and fundamentally limits inference throughput. This limitation is especially acute in physical AI scenarios such as robotics and autonomous driving, where VLMs are deployed on edge devices at batch size one, making AR decoding memory-bandwidth-bound and leaving hardware parallelism underutilized. While block-wise discrete diffusion has shown promise for parallel text generation, extending it to VLMs remains challenging due to the need to jointly handle continuous visual representations and discrete text tokens while preserving pretrained multimodal capabilities. We present Fast-dVLM, a block-diffusion-based VLM that enables KV-cache-compatible parallel decoding and speculative block decoding for inference acceleration. We systematically compare two AR-to-diffusion conversion strategies: a two-stage approach that first adapts the LLM backbone with text-only diffusion fine-tuning before multimodal training, and a direct approach that converts the full AR VLM in one stage. Under comparable training budgets, direct conversion proves substantially more efficient by leveraging the already multimodally aligned VLM; we therefore adopt it as our recommended recipe. We introduce a suite of multimodal diffusion adaptations, block size annealing, causal context attention, auto-truncation masking, and vision efficient concatenation, that collectively enable effective block diffusion in the VLM setting. Extensive experiments across 11 multimodal benchmarks show Fast-dVLM matches its autoregressive counterpart in generation quality. With SGLang integration and FP8 quantization, Fast-dVLM achieves over 6x end-to-end inference speedup over the AR baseline.

cs.CL↗

arXiv2025

Efficient-DLM: From Autoregressive to Diffusion Language Models, and Beyond in Speed

Diffusion language models (dLMs) have emerged as a promising paradigm that enables parallel, non-autoregressive generation, but their learning efficiency lags behind that of autoregressive (AR) language models when trained from scratch. To this end, we study AR-to-dLM conversion to transform pretrained AR models into efficient dLMs that excel in speed while preserving AR models' task accuracy. We achieve this by identifying limitations in the attention patterns and objectives of existing AR-to-dLM methods and then proposing principles and methodologies for more effective AR-to-dLM conversion. Specifically, we first systematically compare different attention patterns and find that maintaining pretrained AR weight distributions is critical for effective AR-to-dLM conversion. As such, we introduce a continuous pretraining scheme with a block-wise attention pattern, which remains causal across blocks while enabling bidirectional modeling within each block. We find that this approach can better preserve pretrained AR models' weight distributions than fully bidirectional modeling, in addition to its known benefit of enabling KV caching, and leads to a win-win in accuracy and efficiency. Second, to mitigate the training-test gap in mask token distributions (uniform vs. highly left-to-right), we propose a position-dependent token masking strategy that assigns higher masking probabilities to later tokens during training to better mimic test-time behavior. Leveraging this framework, we conduct extensive studies of dLMs' attention patterns, training dynamics, and other design choices, providing actionable insights into scalable AR-to-dLM conversion. These studies lead to the Efficient-DLM family, which outperforms state-of-the-art AR models and dLMs, e.g., our Efficient-DLM 8B achieves +5.4%/+2.7% higher accuracy with 4.5x/2.7x higher throughput compared to Dream 7B and Qwen3 4B, respectively.

cs.CL↗

arXiv2025

ToolOrchestra: Elevating Intelligence via Efficient Model and Tool Orchestration

Large language models are powerful generalists, yet solving deep and complex problems such as those of the Humanity's Last Exam (HLE) remains both conceptually challenging and computationally expensive. We show that small orchestrators managing other models and a variety of tools can both push the upper bound of intelligence and improve efficiency in solving difficult agentic tasks. We introduce ToolOrchestra, a method for training small orchestrators that coordinate intelligent tools. ToolOrchestra explicitly uses reinforcement learning with outcome-, efficiency-, and user-preference-aware rewards. Using ToolOrchestra, we produce Orchestrator, an 8B model that achieves higher accuracy at lower cost than previous tool-use agents while aligning with user preferences on which tools are to be used for a given query. On HLE, Orchestrator achieves a score of 37.1%, outperforming GPT-5 (35.1%) while being 2.5x more efficient. On tau2-Bench and FRAMES, Orchestrator surpasses GPT-5 by a wide margin while using only about 30% of the cost. Extensive analysis shows that Orchestrator achieves the best trade-off between performance and cost under multiple metrics, and generalizes robustly to unseen tools. These results demonstrate that composing diverse tools with a lightweight orchestration model is both more efficient and more effective than existing methods, paving the way for practical and scalable tool-augmented reasoning systems.

cs.CL↗

arXiv2025

Nemotron-Flash: Towards Latency-Optimal Hybrid Small Language Models

Efficient deployment of small language models (SLMs) is essential for numerous real-world applications with stringent latency constraints. While previous work on SLM design has primarily focused on reducing the number of parameters to achieve parameter-optimal SLMs, parameter efficiency does not necessarily translate into proportional real-device speed-ups. This work aims to identify the key determinants of SLMs' real-device latency and offer generalizable principles and methodologies for SLM design and training when real-device latency is the primary consideration. Specifically, we identify two central architectural factors: depth-width ratios and operator choices. The former is crucial for small-batch-size latency, while the latter affects both latency and large-batch-size throughput. In light of this, we first study latency-optimal depth-width ratios, with the key finding that although deep-thin models generally achieve better accuracy under the same parameter budget, they may not lie on the accuracy-latency trade-off frontier. Next, we explore emerging efficient attention alternatives to evaluate their potential as candidate building operators. Using the identified promising operators, we construct an evolutionary search framework to automatically discover latency-optimal combinations of these operators within hybrid SLMs, thereby advancing the accuracy-latency frontier. In addition to architectural improvements, we further enhance SLM training using a weight normalization technique that enables more effective weight updates and improves final convergence. Combining these methods, we introduce a new family of hybrid SLMs, called Nemotron-Flash, which significantly advances the accuracy-efficiency frontier of state-of-the-art SLMs, e.g., achieving over +5.5% average accuracy, 1.3x/1.9x lower latency, and 18.7x/45.6x higher throughput compared to Qwen3-1.7B/0.6B, respectively.

cs.LG↗

arXiv2025

TiDAR: Think in Diffusion, Talk in Autoregression

Diffusion language models hold the promise of fast parallel generation, while autoregressive (AR) models typically excel in quality due to their causal structure aligning naturally with language modeling. This raises a fundamental question: can we achieve a synergy with high throughput, higher GPU utilization, and AR level quality? Existing methods fail to effectively balance these two aspects, either prioritizing AR using a weaker model for sequential drafting (speculative decoding), leading to lower drafting efficiency, or using some form of left-to-right (AR-like) decoding logic for diffusion, which still suffers from quality degradation and forfeits its potential parallelizability. We introduce TiDAR, a sequence-level hybrid architecture that drafts tokens (Thinking) in Diffusion and samples final outputs (Talking) AutoRegressively - all within a single forward pass using specially designed structured attention masks. This design exploits the free GPU compute density, achieving a strong balance between drafting and verification capacity. Moreover, TiDAR is designed to be serving-friendly (low overhead) as a standalone model. We extensively evaluate TiDAR against AR models, speculative decoding, and diffusion variants across generative and likelihood tasks at 1.5B and 8B scales. Thanks to the parallel drafting and sampling as well as exact KV cache support, TiDAR outperforms speculative decoding in measured throughput and surpasses diffusion models like Dream and Llada in both efficiency and quality. Most notably, TiDAR is the first architecture to close the quality gap with AR models while delivering 4.71x to 5.91x more tokens per second.

cs.CL↗

arXiv2025

Fast-dLLM v2: Efficient Block-Diffusion LLM

Autoregressive (AR) large language models (LLMs) have achieved remarkable performance across a wide range of natural language tasks, yet their inherent sequential decoding limits inference efficiency. In this work, we propose Fast-dLLM v2, a carefully designed block diffusion language model (dLLM) that efficiently adapts pretrained AR models into dLLMs for parallel text generation, requiring only approximately 1B tokens of fine-tuning. This represents a 500x reduction in training data compared to full-attention diffusion LLMs such as Dream (580B tokens), while preserving the original model's performance. Our approach introduces a novel training recipe that combines a block diffusion mechanism with a complementary attention mask, enabling blockwise bidirectional context modeling without sacrificing AR training objectives. To further accelerate decoding, we design a hierarchical caching mechanism: a block-level cache that stores historical context representations across blocks, and a sub-block cache that enables efficient parallel generation within partially decoded blocks. Coupled with our parallel decoding pipeline, Fast-dLLM v2 achieves up to 2.5x speedup over standard AR decoding without compromising generation quality. Extensive experiments across diverse benchmarks demonstrate that Fast-dLLM v2 matches or surpasses AR baselines in accuracy, while delivering state-of-the-art efficiency among dLLMs - marking a significant step toward the practical deployment of fast and accurate LLMs. Code and model will be publicly released.

cs.CL↗

arXiv2025

Fewer Denoising Steps or Cheaper Per-Step Inference: Towards Compute-Optimal Diffusion Model Deployment

Diffusion models have shown remarkable success across generative tasks, yet their high computational demands challenge deployment on resource-limited platforms. This paper investigates a critical question for compute-optimal diffusion model deployment: Under a post-training setting without fine-tuning, is it more effective to reduce the number of denoising steps or to use a cheaper per-step inference? Intuitively, reducing the number of denoising steps increases the variability of the distributions across steps, making the model more sensitive to compression. In contrast, keeping more denoising steps makes the differences smaller, preserving redundancy, and making post-training compression more feasible. To systematically examine this, we propose PostDiff, a training-free framework for accelerating pre-trained diffusion models by reducing redundancy at both the input level and module level in a post-training manner. At the input level, we propose a mixed-resolution denoising scheme based on the insight that reducing generation resolution in early denoising steps can enhance low-frequency components and improve final generation fidelity. At the module level, we employ a hybrid module caching strategy to reuse computations across denoising steps. Extensive experiments and ablation studies demonstrate that (1) PostDiff can significantly improve the fidelity-efficiency trade-off of state-of-the-art diffusion models, and (2) to boost efficiency while maintaining decent generation fidelity, reducing per-step inference cost is often more effective than reducing the number of denoising steps. Our code is available at https://github.com/GATECH-EIC/PostDiff.

cs.CV↗

arXiv2025

LaCache: Ladder-Shaped KV Caching for Efficient Long-Context Modeling of Large Language Models

Recent advancements in Large Language Models (LLMs) have spurred interest in numerous applications requiring robust long-range capabilities, essential for processing extensive input contexts and continuously generating extended outputs. As sequence lengths increase, the number of Key-Value (KV) pairs in LLMs escalates, creating a significant efficiency bottleneck. In this paper, we propose a new KV cache optimization paradigm called LaCache, a training-free method for efficient and accurate generative inference of LLMs. LaCache enables LLMs to simultaneously address both of the critical challenges in long-range modeling: robust long-range capabilities and continuous generation without running out-of-memory (OOM). Specifically, LaCache integrates two key innovations: (1) a ladder-shaped KV cache pattern that stores KV pairs not only sequentially (left-to-right within each layer) but also across layers (from shallow to deep), providing an extended span for capturing long-range dependencies under a fixed storage budget, thereby boosting long-range capabilities; and (2) an iterative compaction mechanism that progressively compresses older caches, freeing up space for new tokens within a fixed cache size. This token distance-based dynamic compression enables more effective continuous generation under constrained cache budgets. Experiments across various tasks, benchmarks, and LLM models consistently validate LaCache's effectiveness in enhancing LLMs' long-range capabilities. Our code is available at https://github.com/GATECH-EIC/LaCache.

cs.LG↗

arXiv2025

Small Language Models are the Future of Agentic AI

Large language models (LLMs) are often praised for exhibiting near-human performance on a wide range of tasks and valued for their ability to hold a general conversation. The rise of agentic AI systems is, however, ushering in a mass of applications in which language models perform a small number of specialized tasks repetitively and with little variation. Here we lay out the position that small language models (SLMs) are sufficiently powerful, inherently more suitable, and necessarily more economical for many invocations in agentic systems, and are therefore the future of agentic AI. Our argumentation is grounded in the current level of capabilities exhibited by SLMs, the common architectures of agentic systems, and the economy of LM deployment. We further argue that in situations where general-purpose conversational abilities are essential, heterogeneous agentic systems (i.e., agents invoking multiple different models) are the natural choice. We discuss the potential barriers for the adoption of SLMs in agentic systems and outline a general LLM-to-SLM agent conversion algorithm. Our position, formulated as a value statement, highlights the significance of the operational and economic impact even a partial shift from LLMs to SLMs is to have on the AI agent industry. We aim to stimulate the discussion on the effective use of AI resources and hope to advance the efforts to lower the costs of AI of the present day. Calling for both contributions to and critique of our position, we commit to publishing all such correspondence at https://research.nvidia.com/labs/lpr/slm-agents.

cs.AI↗

arXiv2025

LongMamba: Enhancing Mamba's Long Context Capabilities via Training-Free Receptive Field Enlargement

State space models (SSMs) have emerged as an efficient alternative to Transformer models for language modeling, offering linear computational complexity and constant memory usage as context length increases. However, despite their efficiency in handling long contexts, recent studies have shown that SSMs, such as Mamba models, generally underperform compared to Transformers in long-context understanding tasks. To address this significant shortfall and achieve both efficient and accurate long-context understanding, we propose LongMamba, a training-free technique that significantly enhances the long-context capabilities of Mamba models. LongMamba builds on our discovery that the hidden channels in Mamba can be categorized into local and global channels based on their receptive field lengths, with global channels primarily responsible for long-context capability. These global channels can become the key bottleneck as the input context lengthens. Specifically, when input lengths largely exceed the training sequence length, global channels exhibit limitations in adaptively extend their receptive fields, leading to Mamba's poor long-context performance. The key idea of LongMamba is to mitigate the hidden state memory decay in these global channels by preventing the accumulation of unimportant tokens in their memory. This is achieved by first identifying critical tokens in the global channels and then applying token filtering to accumulate only those critical tokens. Through extensive benchmarking across synthetic and real-world long-context scenarios, LongMamba sets a new standard for Mamba's long-context performance, significantly extending its operational range without requiring additional training. Our code is available at https://github.com/GATECH-EIC/LongMamba.

cs.CL↗

arXiv2025

Nemotron-CLIMB: CLustering-based Iterative Data Mixture Bootstrapping for Language Model Pre-training

Pre-training datasets are typically collected from web content and lack inherent domain divisions. For instance, widely used datasets like Common Crawl do not include explicit domain labels, while manually curating labeled datasets such as The Pile is labor-intensive. Consequently, identifying an optimal pre-training data mixture remains a challenging problem, despite its significant benefits for pre-training performance. To address these challenges, we propose CLustering-based Iterative Data Mixture Bootstrapping (Nemotron-CLIMB), an automated framework that discovers, evaluates, and refines data mixtures in a pre-training setting. Specifically, Nemotron-CLIMB embeds and clusters large-scale datasets in a semantic space and then iteratively searches for optimal mixtures using a smaller proxy model and a predictor. When continuously trained on 400B tokens with this mixture, our 1B model exceeds the state-of-the-art Llama-3.2-1B by 2.0%. Moreover, we observe that optimizing for a specific domain (e.g., Social Sciences) yields a 5% improvement over random sampling. Finally, we introduce Nemotron-ClimbLab, a filtered 1.2-trillion-token corpus with 20 clusters as a research playground, and Nemotron-ClimbMix, a compact yet powerful 400-billion-token dataset designed for efficient pre-training that delivers superior performance under an equal token budget. We analyze the final data mixture, elucidating the characteristics of an optimal data mixture. Our data is available at: https://research.nvidia.com/labs/lpr/climb/

cs.CL↗

arXiv2025

Gaussian Blending Unit: An Edge GPU Plug-in for Real-Time Gaussian-Based Rendering in AR/VR

The rapidly advancing field of Augmented and Virtual Reality (AR/VR) demands real-time, photorealistic rendering on resource-constrained platforms. 3D Gaussian Splatting, delivering state-of-the-art (SOTA) performance in rendering efficiency and quality, has emerged as a promising solution across a broad spectrum of AR/VR applications. However, despite its effectiveness on high-end GPUs, it struggles on edge systems like the Jetson Orin NX Edge GPU, achieving only 7-17 FPS -- well below the over 60 FPS standard required for truly immersive AR/VR experiences. Addressing this challenge, we perform a comprehensive analysis of Gaussian-based AR/VR applications and identify the Gaussian Blending Stage, which intensively calculates each Gaussian's contribution at every pixel, as the primary bottleneck. In response, we propose a Gaussian Blending Unit (GBU), an edge GPU plug-in module for real-time rendering in AR/VR applications. Notably, our GBU can be seamlessly integrated into conventional edge GPUs and collaboratively supports a wide range of AR/VR applications. Specifically, GBU incorporates an intra-row sequential shading (IRSS) dataflow that shades each row of pixels sequentially from left to right, utilizing a two-step coordinate transformation. When directly deployed on a GPU, the proposed dataflow achieved a non-trivial 1.72x speedup on real-world static scenes, though still falls short of real-time rendering performance. Recognizing the limited compute utilization in the GPU-based implementation, GBU enhances rendering speed with a dedicated rendering engine that balances the workload across rows by aggregating computations from multiple Gaussians. Experiments across representative AR/VR applications demonstrate that our GBU provides a unified solution for on-device real-time rendering while maintaining SOTA rendering quality.

cs.GR↗

arXiv2024

Hymba: A Hybrid-head Architecture for Small Language Models

We propose Hymba, a family of small language models featuring a hybrid-head parallel architecture that integrates transformer attention mechanisms with state space models (SSMs) for enhanced efficiency. Attention heads provide high-resolution recall, while SSM heads enable efficient context summarization. Additionally, we introduce learnable meta tokens that are prepended to prompts, storing critical information and alleviating the "forced-to-attend" burden associated with attention mechanisms. This model is further optimized by incorporating cross-layer key-value (KV) sharing and partial sliding window attention, resulting in a compact cache size. During development, we conducted a controlled study comparing various architectures under identical settings and observed significant advantages of our proposed architecture. Notably, Hymba achieves state-of-the-art results for small LMs: Our Hymba-1.5B-Base model surpasses all sub-2B public models in performance and even outperforms Llama-3.2-3B with 1.32% higher average accuracy, an 11.67x cache size reduction, and 3.49x throughput.

cs.CL↗

arXiv2024

AmoebaLLM: Constructing Any-Shape Large Language Models for Efficient and Instant Deployment

Motivated by the transformative capabilities of large language models (LLMs) across various natural language tasks, there has been a growing demand to deploy these models effectively across diverse real-world applications and platforms. However, the challenge of efficiently deploying LLMs has become increasingly pronounced due to the varying application-specific performance requirements and the rapid evolution of computational platforms, which feature diverse resource constraints and deployment flows. These varying requirements necessitate LLMs that can adapt their structures (depth and width) for optimal efficiency across different platforms and application specifications. To address this critical gap, we propose AmoebaLLM, a novel framework designed to enable the instant derivation of LLM subnets of arbitrary shapes, which achieve the accuracy-efficiency frontier and can be extracted immediately after a one-time fine-tuning. In this way, AmoebaLLM significantly facilitates rapid deployment tailored to various platforms and applications. Specifically, AmoebaLLM integrates three innovative components: (1) a knowledge-preserving subnet selection strategy that features a dynamic-programming approach for depth shrinking and an importance-driven method for width shrinking; (2) a shape-aware mixture of LoRAs to mitigate gradient conflicts among subnets during fine-tuning; and (3) an in-place distillation scheme with loss-magnitude balancing as the fine-tuning objective. Extensive experiments validate that AmoebaLLM not only sets new standards in LLM adaptability but also successfully delivers subnets that achieve state-of-the-art trade-offs between accuracy and efficiency.

cs.LG↗