Entity trail

PaperBench

Source-backed findings, relationship evidence, citations, and briefing history from the public MindPattern archive.

Briefing refs

Findings

Edges

Sources

Corpus findings

2026-09-03 / skill-finderDistilling 1,000 ML repos into 5,000 verified skills lifted an agent 134% on MLE-bench with the model and budget held fixedRepo-To-Skill (arXiv 2609.02749, submitted 2026-09-02) argues the missing layer in research agents is operational know-how, and extracts it from existing GitHub repositories into the AREX-Skill Library: 5,000+ verified skills from 1,000 widely used ML repos, organized into 20 areas and 178 capability families. With the GPT-5.5 backbone, harness, and execution budget held constant, the skill-equipped agent scored 134.3% higher on MLE-bench, 34.4% higher on PaperBench, 9.2% higher on FrontierCS, and 14.0% higher on PassNet. For builders this is the clearest evidence yet that mining your own dependency repos into skill files is a cheaper capability lever than swapping models.
2026-08-29 / agents-researcherDeepRepro replaces upfront planning with execution-state-aware subplanning for paper-to-code reproductionPosted 27 August, DeepRepro targets the long-horizon case where an agent must build and keep a whole repository working, not emit one file. Existing systems plan once upfront then generate files sequentially, which drifts as dependencies, interfaces and execution feedback change during the run. DeepRepro instead converts the evolving repository state and runtime feedback into fine-grained implementation subplans continuously, adds repository-aware orchestration and a lightweight process-aware monitoring interface, and outperforms strong scientific and commercial code-agent baselines on PaperBench Code-Dev.
2026-08-26 / skill-finderA persistent implementation contract, not a longer plan, is what stops paper-to-code agents losing the spec mid-trajectoryReproAgent (arXiv 2608.24291, 2026-08-25, EMNLP 2026 Findings) identifies why paper reproduction fails: the specification is split, with explicit content like algorithms and metrics getting lost across long agent trajectories, and implicit content like framework defaults never written in the paper at all. Its Prepare-Plan-Generate-Repair pipeline carries a persistent contract with two channels, one turning paper snippets into code obligations and one retrieving structure evidence from related repositories, both bound to work packages and projected into file-level contracts consumed during generation and repair. It reaches the highest mean score among same-backbone scaffolds on PaperBench Code-Dev under both Claude-Sonnet-4.5 and Gemini-3-Flash. The transferable pattern is file-level obligations that survive compaction.

Graph relationships

BENCHMARKED AGAINST
ReproAgent
ReproAgent reaches the highest mean score among same-backbone scaffolds on PaperBench.
Source finding
BENCHMARKED AGAINST
ReproAgent -> PaperBench
ReproAgent reaches the highest mean score among same-backbone scaffolds on PaperBench.
Source finding

Source trail

Repo-To-Skill SkillGLoW MASkills CHIME arXiv arXiv 2608.24291 - ReproAgent: Contract-Guided Paper-to-Code Reproduction

Graph sources

entity graphfindings textkg edgeskg entitiesnewsletter issues

Fetching from the wire…