Cover Image

Hidden States as Early Signals: Step-level Trace Evaluation and Pruning for Efficient Test-Time Scaling

Fuente: arXiv

Saved in:

Bibliographic Details
Main Authors:	Liang, Zhixiang, Huang, Beichen, Wang, Zheng, Zhang, Minjia
Format:	Preprint
Published:	2026
Subjects:	Machine Learning
Online Access:	Acceder al recurso
Tags:	Add Tag No Tags, Be the first to tag this record!

Similar Items

MiLo: Efficient Quantized MoE Inference with Mixture of Low-Rank Compensators
by: Huang, Beichen, et al.
Published: (2025)

PuzzleMoE: Efficient Compression of Large Mixture-of-Experts Models via Sparse Expert Merging and Bit-packed inference
by: Zhao, Yushu, et al.
Published: (2025)

Revisiting the Effectiveness of LLM Pruning for Test-Time Scaling
by: Monjur, Ocean, et al.
Published: (2026)

Inference-Time Chain-of-Thought Pruning with Latent Informativeness Signals
by: Li, Sophie, et al.
Published: (2025)

Learning to (Learn at Test Time): RNNs with Expressive Hidden States
by: Sun, Yu, et al.
Published: (2024)

Downstream-Pretext Domain Knowledge Traceback for Active Learning
by: Zhang, Beichen, et al.
Published: (2024)

Efficient Test-Time Scaling via Self-Calibration
by: Huang, Chengsong, et al.
Published: (2025)

Towards Efficient Automatic Self-Pruning of Large Language Models
by: Huang, Weizhong, et al.
Published: (2025)

Test-Time Scaling in Diffusion LLMs via Hidden Semi-Autoregressive Experts
by: Lee, Jihoon, et al.
Published: (2025)

Critique to Verify: Accurate and Honest Test-Time Scaling with RL-Trained Verifiers
by: Yang, Zhicheng, et al.
Published: (2025)

When Fewer Layers Break More Chains: Layer Pruning Harms Test-Time Scaling in LLMs
by: Wang, Keyu, et al.
Published: (2025)

Evaluating Judges as Evaluators: The JETTS Benchmark of LLM-as-Judges as Test-Time Scaling Evaluators
by: Zhou, Yilun, et al.
Published: (2025)

VoltanaLLM: Feedback-Driven Frequency Control and State-Space Routing for Energy-Efficient LLM Serving
by: Yu, Jiahuan, et al.
Published: (2025)

FLOP-Efficient Training: Early Stopping Based on Test-Time Compute Awareness
by: Amer, Hossam, et al.
Published: (2026)

EfficientLLM: Scalable Pruning-Aware Pretraining for Architecture-Agnostic Edge Language Models
by: Xing, Xingrun, et al.
Published: (2025)

Cut Your Losses! Learning to Prune Paths Early for Efficient Parallel Reasoning
by: Bi, Jiaxi, et al.
Published: (2026)

State Tuning: State-based Test-Time Scaling on RWKV-7
by: Xiao, Liu, et al.
Published: (2025)

Beyond What Seems Necessary: Hidden Gains from Scaling Training-Time Reasoning Length under Outcome Supervision
by: Xue, Yihao, et al.
Published: (2026)

Plan and Budget: Effective and Efficient Test-Time Scaling on Reasoning Large Language Models
by: Lin, Junhong, et al.
Published: (2025)

Early Detection of Misinformation for Infodemic Management: A Domain Adaptation Approach
by: Mao, Minjia, et al.
Published: (2024)

Every Rollout Counts: Optimal Resource Allocation for Efficient Test-Time Scaling
by: Wang, Xinglin, et al.
Published: (2025)

PrunePEFT: Iterative Hybrid Pruning for Parameter-Efficient Fine-tuning of LLMs
by: Yu, Tongzhou, et al.
Published: (2025)

Do Not Waste Your Rollouts: Recycling Search Experience for Efficient Test-Time Scaling
by: Wang, Xinglin, et al.
Published: (2026)

SparseSSM: Efficient Selective Structured State Space Models Can Be Pruned in One-Shot
by: Tuo, Kaiwen, et al.
Published: (2025)

S2O: Early Stopping for Sparse Attention via Online Permutation
by: Zhang, Yu, et al.
Published: (2026)

A Fine Evaluation Method for Cube Copying Test for Early Detection of Alzheimer's Disease
by: Jiang, Xinyu, et al.
Published: (2025)

Test-Time Scaling with Reflective Generative Model
by: Wang, Zixiao, et al.
Published: (2025)

SWE-Replay: Efficient Test-Time Scaling for Software Engineering Agents
by: Ding, Yifeng, et al.
Published: (2026)

Beyond the Frontier: Stochastic Backtracking for Efficient Test-Time Scaling
by: Tran, Dao, et al.
Published: (2026)

MetaScale: Test-Time Scaling with Evolving Meta-Thoughts
by: Liu, Qin, et al.
Published: (2025)

SuperOffload: Unleashing the Power of Large-Scale LLM Training on Superchips
by: Lian, Xinyu, et al.
Published: (2025)

FastForward Pruning: Efficient LLM Pruning via Single-Step Reinforcement Learning
by: Yuan, Xin, et al.
Published: (2025)

Kinetics: Rethinking Test-Time Scaling Laws
by: Sadhukhan, Ranajoy, et al.
Published: (2025)

CipherPrune: Efficient and Scalable Private Transformer Inference
by: Zhang, Yancheng, et al.
Published: (2025)

PaCoRe: Learning to Scale Test-Time Compute with Parallel Coordinated Reasoning
by: Hu, Jingcheng, et al.
Published: (2026)

TraceNAS: Zero-shot LLM Pruning via Gradient Trace Correlation
by: Malettira, Prajna G., et al.
Published: (2026)

EPSD: Early Pruning with Self-Distillation for Efficient Model Compression
by: Chen, Dong, et al.
Published: (2024)

DART-ing Through the Drift: Dynamic Tracing of Knowledge Neurons for Adaptive Inference-Time Pruning
by: Tyagi, Abhishek, et al.
Published: (2026)

Pyramidal Hidden Markov Model For Multivariate Time Series Forecasting
by: Huang, YeXin
Published: (2023)

SalienTime: User-driven Selection of Salient Time Steps for Large-Scale Geospatial Data Visualization
by: Chen, Juntong, et al.
Published: (2024)