简体中文 · English
Dr.Credit: Rubric-Grounded Process Credit Assignment for Deep Research Agents¶
L1 core mechanism diagnostic. Incremental rubric support is computed against history. Rubrics and evidence scores are fixtures; no gold answer is read.
Paper information¶
| Field | Details |
|---|---|
| Paper | Dr.Credit: Rubric-Grounded Process Credit Assignment for Deep Research Agents |
| Institution | University of Chinese Academy of Sciences (first author affiliation) |
| First public release | 2026-09-28 (arXiv v1) |
| Original source code | No public author repository found as of 2026-09-30. |
| Adapter/method | dr-credit |
| Local reproduction code | src/auto_research/agent_research/latest_20260930_closure.py |
Original paper summary¶
Background and main changes¶
Compare a tool result with previously accepted support for each rubric and credit only new or partial support. Combine it with outcome advantage so repeated evidence does not earn repeated credit.
flowchart LR
I[Public input and current state] --> M[Incremental rubric credit]
M --> A[Auditable intermediates]
A --> O[Output or state update]Key figure from the paper¶
Figure 2: Core architecture, modules and connections. From the paper; copyright belongs to the authors. Click for the source.
Core mechanism¶
The local reference kernel retains the defining gates, masks, recursion, credit weights or promotion criteria. Tests check gradient isolation, type boundaries and determinism.
Offline and online results¶
Paper benchmarks, speeds and training curves are author results only. The local three-seed mini-suite checks mechanisms and invariants, not paper-scale performance or online gains.
Local reproduction¶
See three-seed diagnostics. diagnostic_only=true excludes these results from formal capability rankings.
Scope¶
The implementation computes incremental support against rubric history using fixture rubrics and evidence scores. It never reads the gold answer.