Skip to content
Demonstrator · items marked Example are invented · what exists today
E2ER

External pipelines · RISE project catalogue

Scholar Loop

A single-maintainer "autonomous AI scientist" that runs the full PhD loop on a single-GPU budget: literature scouting (arXiv + OpenAlex, citation-ranked), grounded hypothesis generation, debate-gated real PyTorch experiments in a smoke -> verify -> full funnel, reflection into a time-decaying skill library, and a number-grounded write-up with self-review — all under a self-stopping budget governor. Sits with the AI-scientist family (Sakana, Agent Laboratory) but leads with deterministic anti-reward-hacking guards rather than scale.

Indexed in RISE · activeConformance with the standard plannedProject site

What it does

Treats the outer loop and its integrity guards as the product: two-phase frozen scoring the experiment code cannot fake, an edit allowlist, a VerifiedRegistry that grounds every number in the draft, and universal predict-then-verify calibration that scores each agent's checkable claims against ground truth — with a bundled adversarial "cheater" engine to prove the guards hold, and the whole eight-agent loop testable deterministically without an API key or GPU.

Focus
end-to-end
Inputs
domain-profile, budget-config
Outputs
paper-draft, review-report, run-ledger, skill-library
Architecture
multi-agent, tool-use, iterative-loop, persistent-memory, debate-consensus
Maintained by
renee-jia
Started
2026

Description

Data model
Discipline
Computer science
Method family
not specified
Design
not specified
Research stage
Literature discoveryLiterature synthesisHypothesesCode generationData analysisDraftingReview
Contributors
renee-jia
Usage
not used in published research yet
Source
RISE project catalogue · projects/landscape · @4c17bae
Record
pipeline:scholar-loop · JSON

Solid tags are declared by the source or mapped from its terms; dashed tags are inferred by a published rule. Hover a tag for its provenance.

Bring it into the standard

A pipeline built outside E2ER can meet the standard by describing its steps as a template, attaching the floor of checks and publishing evaluation records. Its authors keep ownership and credit.

Other pipelines in Computer science