MovieLens 25M
PulseAugur coverage of MovieLens 25M — every cluster mentioning MovieLens 25M across labs, papers, and developer communities, ranked by signal.
-
New diagnostic method audits recommender system control knobs
Researchers have developed a new diagnostic method to audit the effectiveness of "return conditioning" in recommender systems that utilize Decision Transformers. This method tests how changes in return-to-go (RTG) token…
-
LLMs Under-Confident in Recommendations, Study Finds
A new study auditing four large language models—Mistral Large, Llama 3.3 70B Instruct, GPT-OSS 120B, and Claude Sonnet 4.6—reveals that these models are systematically under-confident when asked to recommend items from …
-
New research details popularity bias impact on AI recommender embeddings
A new research paper published on arXiv explores the impact of popularity-biased feedback on collaborative filtering embeddings. The study introduces a centered-covariance theorem to analyze how this bias reshapes user …
-
Withdrawn paper reveals aggregation distorts user behavior models
A research paper, now withdrawn, explored how data aggregation can distort parametric models of user behavior, a phenomenon termed Simpson's Paradox in behavioral curves. The study, which analyzed data from Goodreads an…
-
New benchmark Loopzero tests recursive collapse warnings
Researchers have developed Loopzero, a new benchmark framework designed to test claims about recursive collapse warnings in complex systems. The framework evaluates telemetry patterns such as rising gain, recursive pers…