Update on Research PSCLS
Summary
The author shares early progress on Leo/PSCLS, an experimental system that learns sequence relationships and improves its story generation and metrics as it is trained on more stories.
Similar Articles
From RLVR to RLSVR (GitHub Repo)
Introduces RLSVR, a task-transformation paradigm that extends RLVR to open-ended tasks via self-verifiable rewards from self-play games, instantiated in SpyRL and Vision-Zero. It improves LLM performance on summarization, creative writing, and math reasoning.
Ran across a site running AI models thru a longford SF fiction test...
A site ran longform speculative-fiction prompts through AI models including Claude Fable 5, publishing the resulting story 'Headwaters' with process notes, raising questions about language becoming training material that people might need to hide.
Detecting LLM-Generated Texts with "Classical" Machine Learning
A developer explores using classical machine learning to detect LLM-generated web fiction, creating an open-source demo and model with ~85% single-sentence accuracy.
@lukaskuhn77: We introduce LeVLJEPA: the first fully non-contrastive end-to-end vision-language pretraining method competitive with C…
LeVLJEPA is the first fully non-contrastive end-to-end vision-language pretraining method, achieving competitive performance with CLIP and SigLIP without negatives, temperature, or momentum encoder. It learns via cross-modal prediction with stop-gradient targets and per-modality distributional regularization, providing stronger dense semantic features for downstream tasks like VLM backbones and semantic segmentation.
@emollick: There is a lot being written about the stylistic tells of AI writing (em-dashes, etc.) but this paper looks at AI narra…
This paper introduces StoryScope, a pipeline that analyzes discourse-level narrative features to distinguish AI-generated fiction from human-written stories. It achieves high accuracy and reveals distinct narrative fingerprints for different LLMs like Claude, GPT, and Gemini.