PtoP / ARCHIVE
Search every paper explained on PtoP. Tags are editorial categories.
The authors argue that decoder-only Transformer LLMs can preserve two text streams at once when the input embeddings are linearly mixed. Instead of collapsing into noise, the model often…
The authors build and fine-tune an open family of language models for K–12 education, but they do not organize the training only by subject or data source. Instead, they design it around…
The paper proposes a way to measure whether an LLM agent makes good long-range decisions while a task is still unfolding. The authors call that ability an agent’s “taste” and build…
Decisions, assumptions and repair
Sound, distance and navigation
Object permanence in video
No papers found. Try another term or tag.