Find a story
Search Spins
Search titles, summaries, and missing voices across published articles — press releases, announcements, and media coverage.
0 results for “training-free”
ConWriter: Transition-Constrained Stateful Long-Form Story Generation with Lightweight Neuro-Symbolic Consistency Control
ConWriter is a new training-free neuro-symbolic framework for long-form story generation that enforces narrative consistency at the scene level using dynamic memory and symbolic state reasoning.
Aug 7, 2026
RubricReviewer: From Direct Critique to Objective and Comprehensive Rubric-Driven Peer Review
RubricReviewer is a new LLM-based peer review framework that explicitly separates rubric generation from review writing to improve comprehensiveness, discriminative quality, and robustness against adversarial attacks on real-world submissions.
Aug 4, 2026
Co-Evolving Graph and Text Memory for Training-Free Multi-Hop Question Answering
A new training-free multi-hop question answering system called Co-E synchronizes graph and text memory to improve reasoning across benchmarks without model retraining.
Jul 28, 2026
Beyond a Global Norm: Personalizing Toxicity Sensitivity in Language Models Without Retraining
Researchers introduced a new framework for personalizing language model toxicity sensitivity without retraining, using inference-time interventions across pre-, in-, and post-decoding stages, revealing trade-offs between alignment accuracy, personalization, and language quality.
Jul 28, 2026
MultAttnAttrib: Training-Free Multimodal Attribution in Long Document Question Answering
Researchers introduced MultAttnAttrib, a training-free method for attributing AI-generated answers to multimodal evidence in long documents, alongside MultAttrEval — the first benchmark dataset for fine-grained multimodal attribution — to address trust and safety gaps in grounded QA systems.
Published Jul 3, 2026 · Analyzed Jul 6, 2026