[KongchangAI]
Model

DeepInstructor

An agent framework for evaluating research ideas based on structured academic experience. It constructs a peer-review experience graph and uses ReAct-paradigm agent retrieval to perform interpretable multi-dimensional assessment of research ideas across novelty, significance, and feasibility.

Timeline (last 90 days)

Sep 22

DeepInstructor is an agent framework that redefines research idea evaluation as a reasoning process over structured academic experience.

Unverified50%
Sep 22

DeepInstructor's experience graph is constructed from 58,607 peer reviews.

Unverified50%
Sep 22

In alignment with human judgment, DeepInstructor achieves a 24.4% improvement in Hit@1.

Unverified50%
Sep 22

DeepInstructor's experience graph is derived from historical reviews and may inherit existing review biases.

Unverified50%
Sep 22

DeepInstructor is a preprint paper published on arXiv.

Unverified50%

All Facts (5)

Source Articles