60 articles tagged with #training-free. AI-curated summaries with sentiment analysis and key takeaways from 50+ sources.
Researchers have developed BWCache, a training-free method that accelerates Diffusion Transformer (DiT) video generation by up to 6× through block-wise feature caching and reuse. The technique exploits computational redundancy in DiT blocks across timesteps while maintaining visual quality, addressing a key bottleneck in real-world AI video generation applications.
Researchers introduce FreeKV, a training-free optimization framework that dramatically improves KV cache retrieval efficiency for large language models with long context windows. The system achieves up to 13x speedup compared to existing methods while maintaining near-lossless accuracy through speculative retrieval and hybrid memory layouts.
Researchers introduce FOCUS, a training-free method that improves Large Vision-Language Models' ability to process multiple images by masking irrelevant images with noise, preventing visual information from different images from becoming entangled in the model's representations.
Researchers introduce DART, a training-free routing framework that dynamically allocates computational thinking budgets in hybrid reasoning models by sampling cheap draft responses and using agreement patterns to decide between direct answers and extended reasoning. The approach achieves significant accuracy improvements on math and code tasks while reducing token consumption by 15-69%, without requiring labeled data or model fine-tuning.
Researchers present Gazer, a training-free framework that uses multimodal large language models to identify and correct semantic errors in autoregressive visual models during image and video generation. The approach operates through diagnostic and correction stages that analyze intermediate generation states and adjust trajectories without requiring additional model training.
Researchers propose MMD Guidance, a training-free method that uses Maximum Mean Discrepancy to align pre-trained diffusion models with target data distributions during inference. The technique enables domain adaptation without retraining, working efficiently in both standard and latent diffusion models while maintaining sample quality.
Researchers introduce GVC-Seg, a training-free 3D instance segmentation method that uses geometric visual correspondence to eliminate confidence bias when combining multiple foundation models. The approach achieves state-of-the-art results on challenging benchmarks while maintaining strong performance in open-vocabulary semantic segmentation tasks.
Researchers present DAVE, a training-free method that enhances diversity in text-to-image generation by attenuating the DC (zero-frequency) component of intermediate Transformer features during early generation stages. The technique addresses the problem of identical outputs from the same prompt without requiring expensive sampling overhead or auxiliary optimization.
EasyLens is a training-free method that enhances medical vision-language models' ability to detect subtle lesions in clinical images without requiring additional model training or adaptation. The approach uses prototype-based reasoning and representation amplification to ensure weak visual cues from lesions aren't lost in global image representations, outperforming existing enhancement methods across multiple medical datasets.
Researchers introduce Dynamic Infilling Anchors (DIA), a training-free method that improves how diffusion large language models generate structured outputs like JSON or reasoning templates. By dynamically adjusting generation length constraints, DIA achieves better format compliance and accuracy on mathematical reasoning benchmarks without requiring model retraining.
Researchers propose a training-free, lightweight framework for scene text recognition that leverages pre-trained models and context-driven understanding to achieve state-of-the-art performance with significantly reduced computational requirements. The approach uses attention-based segmentation and semantic evaluation to enable faster inference suitable for real-time deployment scenarios.
Researchers introduce DOA (Decoder-Only Attention), a training-free method that enables simultaneous speech-to-text translation using decoder-only SpeechLLMs by extracting alignment signals from self-attention mechanisms. The approach achieves low-latency, long-form translation quality comparable to offline decoding without requiring model retraining.
Researchers propose DLM-SWAI, a training-free method for steering diffusion language models toward desired outputs by biasing token distributions during iterative denoising. The approach enables controllable text generation for style and safety applications without retraining or auxiliary models, addressing a gap in control methods for diffusion-based language generation.
Researchers demonstrate that identity-preserved image generation using FLUX can be accelerated 5.9x by replacing the standard diffusion backbone with a distilled version, without retraining the identity adapter. Analysis reveals identity fidelity stabilizes within 4-8 steps while later steps primarily refine visual details, enabling efficient personalized generation at deployment.
Researchers present InpaintSLat, a training-free method for 3D inpainting that optimizes initial noise in structured 3D latent diffusion models. The approach leverages backpropagation approximation and spectral parameterization to improve geometric stability and contextual consistency, outperforming existing training-free baselines without requiring model retraining.
Researchers introduce MODIX, a training-free framework that dynamically optimizes how Vision-Language Models allocate attention across multimodal inputs by adjusting positional encoding based on information density rather than uniform token assignment. The approach improves reasoning performance without modifying model parameters, suggesting positional encoding should be treated as an adaptive resource in multimodal transformer architectures.
KITE is a training-free system that converts long robot execution videos into compact, interpretable tokens for vision-language models to analyze robot failures. The approach combines keyframe extraction, open-vocabulary detection, and bird's-eye-view spatial representations to enable failure detection, identification, localization, and correction without requiring model fine-tuning.
Researchers introduce VLM-UnBench, the first benchmark for evaluating training-free visual concept unlearning in Vision Language Models. The study reveals that realistic prompts fail to genuinely remove sensitive or copyrighted visual concepts, with meaningful suppression only occurring under oracle conditions that explicitly disclose target concepts.
Researchers developed plan conditioning, a training-free method that significantly improves diffusion language model reasoning by prepending short natural-language plans from autoregressive models. The technique improved performance by 11.6 percentage points on math problems and 12.8 points on coding tasks, bringing diffusion models to competitive levels with autoregressive models.
Researchers developed training-free model steering techniques to improve reasoning in large audio-language models (LALMs) through chain-of-thought prompting. The approach achieved up to 4.4% accuracy gains and demonstrated cross-modal transfer where text-derived steering vectors can effectively guide speech-based reasoning.
Researchers propose Global Evolutionary Refined Steering (GER-steer), a new training-free framework for controlling Large Language Models without fine-tuning costs. The method addresses issues with existing activation engineering approaches by using geometric stability to improve steering vector accuracy and reduce noise.
Researchers introduce CoE, a training-free multimodal summarization framework that uses a Chain-of-Events approach with Hierarchical Event Graph to better understand and summarize content across videos, transcripts, and images. The system achieves significant performance improvements over existing methods, showing average gains of +3.04 ROUGE, +9.51 CIDEr, and +1.88 BERTScore across eight datasets.
Researchers developed MAP (Map-Level Attention Processing), a training-free method to reduce hallucinations in Large Vision-Language Models by treating hidden states as 2D semantic maps. The approach uses attention-based operations to better leverage factual information and improve consistency between generated text and visual inputs.
Researchers propose a training-free paradigm for empowering Vision-Language Models with multi-modal search capabilities through cross-modal model merging. The approach uses Optimal Brain Merging (OBM) to combine text-based search agents with base VLMs without requiring expensive supervised training or reinforcement learning.
AdaFocus is a new training-free framework for adaptive visual reasoning in Multimodal Large Language Models that addresses perceptual redundancy and spatial attention issues. The system uses a two-stage pipeline with confidence-based cropping decisions and semantic-guided localization, achieving 4x faster inference than existing methods while improving accuracy.
