五款 Qwen3.8-27B 模型使用 GVS5H Harness 與 Fable 5 相媲美
Five Qwen3.8-27B models match Fable 5 Using GVS5H Harness
GVS5H 是一種零訓練自組織方法，利用共享檔案系統協調多個模型，能在 LiveCodeBench 上達到接近或超過 Fable 5 的編碼準確率。
- 核心機制：使用 ledger‑based zero‑shot self‑orchestration，模型在推理時拆解問題並共享計畫、筆記與解答。
零訓練自組織方法在 LiveCodeBench 上達到前沿編碼精度，顯示可在成本大幅下降下提升實務效能。
GVS5H: Gorilla vs. 5 Humans
Results
Opus 5 ran on an earlier version of the scaffold, in a single pass served through OpenRouter rather than five on a pinned backend.
Abstract. Frontier coding performance is typically bought with larger proprietary models at high cost. We introduce ledger-based zero-shot self-orchestration, a training-free method in which fresh instances of one model decompose problems and coordinate through a shared filesystem holding a plan, notes and current solution. Across eleven open and closed-weight models on the 100 latest hard LiveCodeBench problems, the method yields gains of up to 25.6 percentage points on pinned backends and offers two routes to frontier-level accuracy. Orchestrated Qwen3.8 Flash Next reaches 93.0% pass@1 against Fable 5's 90.4% at 9% of the cost, and locally served, open-weight Qwen3.8-27B rises from 66.8% to 92.4%. Gains are not universal: some models are unchanged or worse. Transcript analysis attributes the gain to decomposition and persistent context. Inference-time organization can approach frontier coding accuracy at a fraction of the cost, or slightly exceed it on self-hostable weights.
Running the code
Needs uv and an API key for the model you want to test.
cd codebase/v2-current export OPENAI_API_KEY=... LCB_RELEASE=release_v6 \ ESCALATION_CLOUD_MAX_TOKENS=128000 \ ESCALATION_CLOUD_TIMEOUT=7200 \ MULTIAGENT_MODEL=openai:gpt-5.6-terra \ uv run --no-project --python 3.12 --with 'datasets<4' --with numpy --with anthropic \ python escalation/run_bench.py --engine multiagent --only lcb --lcb 100 --parallel 8
--engine multiagentruns the manager;--engine singleis the one-call baseline.- Other models:
anthropic:<model>,dashscope:<model>,openrouter:<model>, each with its own*_API_KEY. - The pass@1 score prints at the end. Results are written to
runs/results.json, workspaces toruns/ws/.
License
Code is under the MIT License. The paper, figures and run data are under CC BY 4.0. The LiveCodeBench fork, the benchmark problem statements and the LaTeX template files keep their own licenses. See NOTICE.md for which license covers which path.