Star Elastic: Many-in-One Reasoning LLMs with Efficient Budget Control

• URL: http://arxiv.org/abs/2605.07182v1
• Date: Fri, 08 May 2026 03:20:36 GMT
• Title: Star Elastic: Many-in-One Reasoning LLMs with Efficient Budget Control
• Abstract summary: We introduce Star Elastic, a novel post-training method for large language models (LLMs)<n>Star Elastic adds N nested submodels to a given parent reasoning model using the compute of one run (N-fold savings) via a single post-training job.<n>Building on the Nemotron Elastic framework, we apply Star Elastic to the NVIDIA Nemotron Nano models.
• Score: 27.041571161298688
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Training a family of large language models (LLMs), either from scratch or via iterative compression, is prohibitively expensive and inefficient, requiring separate training runs for each model in the family. In this paper, we introduce Star Elastic, a novel LLM post-training method that adds N nested submodels to a given parent reasoning model using the compute of one run (N-fold savings) via a single post-training job. Beyond reducing training costs, Star Elastic also addresses a fundamental limitation of efficient reasoning: the rigidity of static architectures, which forces the allocation of constant resources regardless of token difficulty. By unlocking elastic budget control, Star Elastic enables a novel inference scheme that uses different submodels for each reasoning phase (thinking and answering). Star Elastic supports (1) nesting along the SSM, embedding channel, MoE, and FFN axes, (2) learning nested submodels via an end-to-end trainable router, and (3) curriculum-based knowledge distillation. Building on the Nemotron Elastic framework, we apply Star Elastic to the NVIDIA Nemotron Nano models, with a particular focus on hybrid Mixture-of-Experts (MoE) architectures: from Nemotron Nano v3 (30B/3.6A), we generate 23B (2.8A) and 12B (2.0A) variants with 160B training tokens. All nested models match or outperform independently trained baselines of comparable size and achieve a 360x reduction versus pretraining from scratch and a 7x reduction over state-of-the-art compression. Crucially, elastic budget control advances the accuracy-latency Pareto frontier, achieving up to 16% higher accuracy and 1.9x lower latency via dynamic per-phase model selection. We further extend Star Elastic to quantized regimes via Quantization-Aware Distillation (QAD), producing nested NVFP4 and FP8 elastic checkpoints that preserve zero-shot slicing while delivering smaller deployment footprints.

Related papers

• Compact SO(3) Equivariant Atomistic Foundation Models via Structural Pruning [4.793059213046564]
  We propose a structural pruning method for SO(3) equivariant atomistic foundation models.<n>The pruning is applied along the channel and order dimensions, with each irreducible representation kept or removed as a complete block.<n>The pruned MACE-MP model outperforms the official from-scratch trained small model on 7 of 9 metrics on the Matbench Discovery leaderboard.
  arXiv  Detail & Related papers  (2026-05-09T11:07:03Z)
• FlexRank: Nested Low-Rank Knowledge Decomposition for Adaptive Model Deployment [20.331469310989956]
  We argue that importance-ordered nested components can be extracted from pretrained models, and selectively activated on the available computational budget.<n>Our approach enables a "train-once, deploy-everywhere" paradigm that offers a graceful trade-off between cost and performance without training from scratch for each budget.
  arXiv  Detail & Related papers  (2026-02-02T19:01:40Z)
• Nemotron Elastic: Towards Efficient Many-in-One Reasoning LLMs [80.72350166388601]
  Nemotron Elastic is a framework for building reasoning-oriented LLMs.<n>It embeds nested submodels within a single parent model.<n>Each of these submodels shares weights with the parent model and can be extracted zero-shot during deployment.
  arXiv  Detail & Related papers  (2025-11-20T18:59:21Z)
• Every Step Evolves: Scaling Reinforcement Learning for Trillion-Scale Thinking Model [100.86587937568832]
  Ring-1T is the first open-source, state-of-the-art thinking model with a trillion-scale parameter.<n>It features 1 trillion total parameters and activates approximately 50 billion per token.
  arXiv  Detail & Related papers  (2025-10-21T17:46:14Z)
• Elastic ViTs from Pretrained Models without Retraining [74.5386166956142]
  Vision foundation models achieve remarkable performance but are only available in a limited set of pre-determined sizes.<n>We introduce SnapViT: Single-shot network approximation for pruned Vision Transformers.<n>Our approach efficiently combines gradient information with cross-network structure correlations, approximated via an evolutionary algorithm.
  arXiv  Detail & Related papers  (2025-10-20T16:15:03Z)
• CAST: Continuous and Differentiable Semi-Structured Sparsity-Aware Training for Large Language Models [27.682531424487564]
  Sparsity-aware training is an effective approach for transforming large language models into hardware-friendly sparse patterns.<n>We propose Continuous Adaptive Sparse Trainer (CAST), a continuous and differentiable sparsity-aware training framework for sparse models.<n>Our results demonstrate significant improvements over previous state-of-the-art methods in both perplexity and zero-shot accuracy with minimal training resources.
  arXiv  Detail & Related papers  (2025-09-30T09:28:47Z)
• Deep Hierarchical Learning with Nested Subspace Networks [53.71337604556311]
  We propose Nested Subspace Networks (NSNs) for large neural networks.<n>NSNs enable a single model to be dynamically and granularly adjusted across a continuous spectrum of compute budgets.<n>We show that NSNs can be surgically applied to pre-trained LLMs and unlock a smooth and predictable compute-performance frontier.
  arXiv  Detail & Related papers  (2025-09-22T15:13:14Z)
• Flexiffusion: Training-Free Segment-Wise Neural Architecture Search for Efficient Diffusion Models [50.260693393896716]
  Diffusion models (DMs) are powerful generative models capable of producing high-fidelity images but constrained by high computational costs.<n>We propose Flexiffusion, a training-free NAS framework that jointly optimize generation schedules and model architectures without modifying pre-trained parameters.<n>Our work pioneers a resource-efficient paradigm for searching high-speed DMs without sacrificing quality.
  arXiv  Detail & Related papers  (2025-06-03T06:02:50Z)
• Balanced and Elastic End-to-end Training of Dynamic LLMs [2.7461964910607097]
  We propose an autonomous dynamic load balancing solution, DynMo, for large-scale distributed training.<n>DynMo provably achieves maximum reduction in workload imbalance and adaptively equalizes compute loads across workers.<n>Compared to static distributed training solutions such as Megatron-LM and DeepSpeed, DynMo accelerates the end-to-end training of dynamic GPT models by up to 1.23x for MoEs, 3.18x for parameter pruning, 2.23x for layer freezing, 4.02x for sparse attention, 4.52x for early exit, and 1.17x for MoDs
  arXiv  Detail & Related papers  (2025-05-20T19:52:57Z)
• Scalable Chain of Thoughts via Elastic Reasoning [61.75753924952059]
  Elastic Reasoning is a novel framework for scalable chain of thoughts.<n>It separates reasoning into two phases--thinking and solution--with independently allocated budgets.<n>Our approach produces more concise and efficient reasoning even in unconstrained settings.
  arXiv  Detail & Related papers  (2025-05-08T15:01:06Z)

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.