#training-free methods

topictraining-free methods

16 papers · 1 filter

cs.CV2026

ViewMind3D: Modular View-Aware Inference for Training-Free 3D-QA

Ping-Kun Chiang, Kun-Ru Wu, Po-han Li +3

ViewMind3D is a training‑free, modular framework that answers 3D questions by selecting relevant views, grounding objects with language cues, encoding spatial context via a bird's‑…

cs.CV2026

Capturing Token Tendencies for Training-Free Token Pruning in Multimodal Large Language Models

Jie Ma, Zhike Qiu, Jie Gao +4

The paper introduces Trend-aware Pruning, a training‑free method that models the temporal dynamics of attention to selectively keep visual tokens that become important in deeper la…

cs.CV2026

LAST: The Last Query Token Guides Visual Token Pruning for Edge-Cloud Collaborative MLLM Inference

Feng Yang, Xinrui Ju, Keyang Zhang +6

The paper introduces LAST, a training‑free method that uses the attention of the last query token to prune visual tokens on edge devices before sending them to a cloud multimodal L…

cs.LG2026

Prox: Training-Free FFN Activation Sparsity via Approximate Intermediate-Channel Salience in LLMs

Jinyi Liu, Wei Chen, Pengyu Chen +4

The paper introduces Prox, a training-free framework that sparsifies feed‑forward network activations in large language models by approximating intermediate‑channel salience, enabl…

cs.CV2026

FreqForcing: Autoregressive Long Video Generation via Spectral Self-Anchoring

Jiatong Li, Leo Liang, Linghe Kong +1

The paper introduces FreqForcing, a training‑free method that uses spectral self‑anchoring to counter low‑frequency energy drift and improve visual stability in autoregressive long…

cs.CV2026

TraceCLIP: Recovering Local Semantics from Patch-to-CLS Contributions

Xinran Liu, Shouqian Shi, Yutong Chen +3

The paper presents TraceCLIP, a training‑free method that extracts patch‑level semantic information from CLIP's CLS attention output to improve zero‑shot dense vision‑language task…