{"record": {"aliases": ["Backtrader-Bench: Benchmarking LLM Agents on Algorithmic Trading with Self-Generated MCQs"], "artifacts": [{"id": "arxiv:2608.11232", "kind": "paper", "url": "https://arxiv.org/abs/2608.11232"}, {"id": "gh:rzhao999/backtrader-bench", "kind": "repo", "url": "https://github.com/rzhao999/Backtrader-Bench"}], "categories": ["Finance", "Language & Knowledge", "cs.CL"], "description": {"en": "An algorithmic-trading benchmark for LLM coding agents. It uses multiple-choice questions derived from backtests, with an independent checker that re-derives every answer, and scores models by accuracy."}, "documents": [{"document_type": "registry_page", "id": "claire_radar:https://arxiv.org/abs/2608.11232", "source": "claire_radar", "source_url": "https://arxiv.org/abs/2608.11232", "title": "Backtrader-Bench"}], "evidence_summary": {"document_count": 1, "model_count": null, "model_count_basis": null}, "key": "claire-radar:2608.11232", "modality": null, "name": "Backtrader-Bench", "openness": {"code_license": null, "data_license": null, "evidence": [], "status": "unknown"}, "provenance": {"admission_policy_version": "2026-09-09.1", "confidence": "Medium", "crawl_bundle": "claire_radar_2026-10-14", "crawled_at": "2026-09-25T04:40:22Z", "data_status": "primary-source-indexed", "display_eligible": "true", "origin_record_id": "bm_backtrader-bench-benchmarking-llm-agents-o_eb6cfacd", "origin_source": "arxiv", "recognition_confidence": "0.55", "record_sha256": "e26a4c875b47e3408132e08ce3354b5a7ee51d82d20caaf235ec52a7072308f9", "relation": "evaluates_only", "review_model": "claude-code-default", "review_state": "ai-reviewed", "reviewed_at": "2026-09-24T20:54:37.532631Z", "source_url": "https://arxiv.org/abs/2608.11232"}, "publisher": null, "released": "2026-07-31", "released_reference": {"basis": "benchmark_release", "source_key": "claire-radar:2608.11232", "source_url": "https://arxiv.org/abs/2608.11232"}, "schema_version": 1, "sizes": [], "slug": "claire-radar-2608-11232", "source": "claire_radar", "source_benchmark_id": "2608.11232", "source_metadata": {"claire_radar": {"adoption30d": null, "aliases": ["Backtrader-Bench: Benchmarking LLM Agents on Algorithmic Trading with Self-Generated MCQs"], "annotation": "Unknown", "applicationDomains": ["Finance"], "area": "Language & Knowledge", "attentionForecast": {"confidence": "Low", "horizon": "7d", "reason": "A niche finance-agent benchmark from a workshop paper, with public code and data.", "score": 40}, "capabilities": [], "confidence": "Medium", "construction": "Transform Existing", "constructionDetail": "Unknown — the indexer does not infer construction details without explicit source evidence.", "copyGeneration": {"generatedAt": "2026-09-24T20:54:37.532631Z", "inputHash": "deef1b66c22efa9af41d7e126fdc0d0310c85a035e6643c2c6cf6eb97f778394", "model": "claude-code-default"}, "curation": {"admissionPolicyVersion": "2026-09-09.1", "canonicalNameEvidence": "Backtrader-Bench: Benchmarking LLM Agents on Algorithmic Trading with Self-Generated MCQs", "canonicalNameSource": "paper_title", "decisionReason": "The benchmark is formally named, uses an accuracy metric on MCQs whose answers are re-derived by a checker, and the paper supplies a public code and data link.", "evaluationEvidence": {"reuse": "Code and data: https://github.com/rzhao999/Backtrader-Bench", "scoring": "Tool-augmented agents reach 90.0% accuracy in a single pass", "task": "generates questions from backtest configurations across five trading strategies, 33 templates, and three difficulty tiers"}, "model": "claude-code-default", "reviewedAt": "2026-09-24T20:54:37.532631Z", "state": "ai-reviewed"}, "dataStatus": "primary-source-indexed", "demo": false, "description": "An algorithmic-trading benchmark for LLM coding agents. It uses multiple-choice questions derived from backtests, with an independent checker that re-derives every answer, and scores models by accuracy.", "displayEligible": true, "domainCuration": {"method": "rules-v1", "state": "auto"}, "evaluationMode": "public_reusable", "evidence": {"reasonCodes": ["coined title prefix ending in Bench or Benchmark", "benchmark release stated in one sentence", "public artifact URL", "named entity is a model, method, framework, or dataset used in evaluation"], "snippet": "We present Backtrader-Bench, a framework with two complementary pipelines."}, "familyId": "bmf_06d1dff1f4fe", "firstSeenAt": "2026-08-19", "heat": null, "id": "bm_backtrader-bench-benchmarking-llm-agents-o_eb6cfacd", "indexedAt": "2026-08-19T11:05:13.395391Z", "industrySectors": ["Financial Services"], "links": {"code": "https://github.com/rzhao999/Backtrader-Bench", "data": null, "paper": "https://arxiv.org/abs/2608.11232", "pdf": "https://arxiv.org/pdf/2608.11232", "project": null, "report": "https://arxiv.org/abs/2608.11232"}, "metrics": [], "motivation": "Evaluating LLM coding agents in algorithmic trading is difficult because static benchmarks risk data contamination and numerical backtest outputs require ground truth from actual code execution.", "name": "Backtrader-Bench", "oneLine": "An algorithmic-trading benchmark for LLM coding agents. It uses multiple-choice questions derived from backtests, with an independent checker that re-derives every answer, and scores models by accuracy.", "paperTitle": "Backtrader-Bench: Benchmarking LLM Agents on Algorithmic Trading with Self-Generated MCQs", "primaryDomain": "Finance", "publication": {"evidence": "Accepted to the FinLLM Workshop at IJCAI 2026. Code and data: https://github.com/rzhao999/Backtrader-Bench", "evidenceLevel": "author-claim", "evidenceUrl": "https://arxiv.org/abs/2608.11232", "source": "arxiv-comments", "status": "acceptance_claimed", "venue": "FinLLM Workshop at IJCAI 2026", "verifiedAt": "2026-08-19T11:05:13.395391Z"}, "publications": [], "readiness": "Runnable", "recognitionConfidence": 0.55, "relation": "evaluates_only", "releasedAt": "2026-07-31", "reviewLifecycle": {"state": "pending"}, "source": {"authors": ["Ruoxi Zhao", "Maziar Raissi"], "categories": ["cs.CL", "cs.AI"], "id": "2608.11232", "title": "Backtrader-Bench: Benchmarking LLM Agents on Algorithmic Trading with Self-Generated MCQs", "type": "arxiv", "url": "https://arxiv.org/abs/2608.11232"}, "sourceUpdatedAt": "2026-07-31T02:37:57Z", "topics": ["cs.CL"], "venueAttempts": [{"decisionRaw": "Accepted to the FinLLM Workshop at IJCAI 2026. Code and data: https://github.com/rzhao999/Backtrader-Bench", "evidence": [{"level": "author-claim", "observedAt": "2026-08-19T11:05:13.395391Z", "rawValue": "Accepted to the FinLLM Workshop at IJCAI 2026. Code and data: https://github.com/rzhao999/Backtrader-Bench", "sourceType": "arxiv-comments", "sourceUrl": "https://arxiv.org/abs/2608.11232"}], "reviewStatus": "accepted", "venueName": "FinLLM Workshop at IJCAI 2026"}], "whyItMatters": "It targets data contamination in finance-agent evaluation by generating questions whose answers are grounded in code execution, and it separates results with and without tool use."}}}, "schema_version": 1, "scores_by_source": {}, "siblings": []}
