Tag
VAKE is a two-stage reinforcement-learning framework that externalizes latent parametric knowledge in LLMs through explicit priming and implicit reasoning, achieving superior performance across multiple benchmarks.
The author argues that while tool use allows smaller language models to perform tasks effectively, larger models remain crucial for speed, reliability, and internalized knowledge, emphasizing the ongoing need for scaling in AI.
This paper introduces NameRank, a method to measure how well LLMs recognize individuals and artifacts from their parametric knowledge, finding that models recognize named projects and methods far better than credentials or contributors.
This paper explores cross-lingual prompting strategies to improve access to parametric knowledge in large language models, demonstrating significant gains in knowledge transfer and factual recall across 17 languages on multilingual benchmarks.
Decoupled Mixture-of-Experts (DMoE) proposes a modular architecture for parametric knowledge injection, decoupling experts and router from the base model to enable efficient auto-regressive inference and mitigate catastrophic forgetting.
ToolSense is an open-source diagnostic framework that generates three benchmarks (realistic retrieval, MCQ probing, QA probing) to audit LLMs' parametric tool knowledge, revealing a knowledge-retrieval dissociation where strong retrieval performance can coexist with poor factual understanding.
This paper investigates whether reinforcement learning can improve the direct recall of parametric knowledge in LLMs beyond reasoning tasks. It demonstrates that RL with binary rewards yields significant gains in factual QA benchmarks by redistributing probability mass to unlock latent knowledge rather than acquiring new facts.