Skip to content
AI-generated research idea🔴Hard🔬Significant ResearchNo close match found

Subgraph-Decomposed Curriculum RL for Hard Repository-Level Software Engineering

Apply curriculum-based reinforcement learning over decomposed dependency subgraphs of failed repository-generation tasks so coding agents learn multi-file software construction instead of discarding unsolved problems.

Time

a couple of months

Compute level

Large cluster

Resources

Large GPU cluster for 320B MoE RL rollouts plus high-throughput CPU container fleet (4 CPUs, 16 GiB RAM per task sandbox as used in SWE-bench Pro).

What the paper used

The paper reports a 320-billion-parameter MoE model, ToolSpace data covering 77,760 tool-call pattern nodes across 2,042 tools and 804 environments, and ML Coding candidate inventories of 2,960 papers and 5,636 dataset sources, but does not state the GPU hardware or training compute budget used for pretraining or post-training (Appendix B only specifies 1-4 CPUs and 8-16 GiB memory per task container during benchmark evaluation).

Prerequisites

Reinforcement learning for code agents, static program analysis/code knowledge graphs, and containerized SWE evaluation harnesses.

Scope

new curriculum synthesis and RL pipeline

The Gap

In Section 3.1.3 (Agentic Coding), the pipeline constructs repository-level generation and patch-level SWE tasks using GitNexus code knowledge graphs and mines multiple rollouts per task, filtering out tasks for which no successful solution is obtained. In Table 2, Sage S2.1 Flash trails leading baselines substantially on repository-level and SWE benchmarks (57.2 vs. 79.9 on NL2Repo-Bench; 60.6 vs. 76.5 on SWE-bench Pro; 67.9 vs. 73.9 on DeepSWE).

“Empirical pass rates inform task filtering, removing some overly easy tasks and tasks for which no successful solution is obtained under the current sampling configuration.”
From the paper, Section 3.1.3, page 8. Matched against the PDF text.

Limitation

Discarding tasks with zero successful rollouts under current sampling budgets starves the model of training signal on precisely the hardest multi-file dependency, interface-evolution, and architectural synthesis problems reflected in NL2Repo-Bench and SWE-bench Pro.

Proposed Solution

Instead of dropping zero-pass repository tasks, use the GitNexus code-structure graph to decompose each failed repository task into an ordered curriculum of verifiable sub-tasks (e.g., core data structures, module interfaces, integration tests). Run multi-turn RL (GRPO) with dense intermediate rewards combining unit-test pass fraction, static interface/type-check compatibility, and regression-free patch execution, gradually merging sub-tasks back into full repository-level synthesis.

The paper’s authors list this direction as future work, so they may already be pursuing it.

Why This Matters

Recovering learning signal from previously unsolved tasks is expected to narrow the 15-22 point gap on repository-level coding benchmarks while improving multi-file compatibility.

How to Get Started

How to test it: Hypothesis: Curriculum RL over subgraph-decomposed zero-pass repository tasks outperforms standard rejection-sampled SFT on long-horizon coding benchmarks. Baseline: Current Sage S2.1 Flash checkpoint evaluated on NL2Repo-Bench (57.2%) and SWE-bench Pro (60.6%); stop if pass rates on held-out NL2Repo-Bench tasks fail to improve by at least 4 percentage points after curriculum RL.

  1. Collect the zero-pass repository-level and SWE tasks filtered out during the Section 3.1.3 rollout collection stage (or replicate zero-pass tasks from NL2Repo-Bench/OpenSWE training splits).
  2. Use GitNexus dependency graphs and AST analysis to split each failed task into staged interface and module-level sub-tasks with runnable unit-test subsets.
  3. Implement partial unit-test and interface-compatibility reward shaping in XTuner over Harbor/OpenHands sandboxes.
  4. Train with GRPO from stage-1 sub-tasks to full-repository tasks and benchmark on NL2Repo-Bench and SWE-bench Pro.

Suggested by AI. Adapt it to your setup.

Has This Been Done?

No close match in 10 results

Works like WEAVE discuss repository-level benchmarking and task decomposition, but none combine GitNexus-based code-structure graph decomposition into ordered verifiable sub-tasks with multi-turn GRPO and dense intermediate rewards for code agents.

We searched OpenAlex and arXiv for “repository level code generation curriculum reinforcement learning dependency graph” and “Subgraph-Decomposed Curriculum RL for Hard Repository-Level Software Engineering” and an AI compared the idea with the top 10 results. A small search can miss relevant work.

Source Paper

Sage S2.1 Flash Technical Report

Kun Cai, Haodong Chen, Guangran Cheng +203 more · 2026

Sage S2.1 Flash is a 320-billion-parameter mixture-of-experts (MoE) model designed to integrate native scientific perception with general-purpose multimodal reasoning and long-horizon agentic execution. It combines dedicated scientific encoders and modality-specific projectors (Sage Sci-Perception) with post-training across large-scale verifiable environments spanning tool use, search, coding, ML engineering, workspace deliverables, and computer use, achieving top-two rankings on many of 33 evaluated benchmarks.

Related Ideas

Is this a good idea?

Share onXLinkedIn