📋 摘要
⭐ 涉及LLM agent可靠性与执行护栏，与可信AI略有交集，但聚焦onchain交易，偏离SE/测试/形式化主线。
Operating-Layer Controls for Onchain Language-Model Agents Under Real Capital
中文
本文研究在真实资本环境下，自主 language-model agents 将用户指令转化为合规工具动作的可靠性问题。研究依托 DX Terminal Pro 平台，进行了为期 21 天的部署：3,505 个由用户出资的 agents 在受限的 onchain 市场中交易真实 ETH。用户通过结构化控件与自然语言策略配置 vault，但仅 agents 可决定常规买卖交易。系统累计产生约 750 万次 agent 调用、约 30 万次 onchain 动作、约 2,000 万美元交易量、5,000+ ETH 部署量及约 700 亿 inference tokens，policy-valid 交易的结算成功率达 99.9%；长期运行的 agents 累积了数千次顺序决策，包含 6,000+ 个 prompt-state-action 循环，形成覆盖用户指令、prompt 渲染、推理、校验、组合状态到结算的大规模 trace。作者指出，可靠性并非源自基础模型本身，而源自模型外围的 operating layer：prompt 编译、typed controls、policy 校验、执行护栏、记忆设计与 trace 级可观测性。预上线测试暴露了纯文本 benchmark 难以衡量的失效模式，包括捏造交易规则、fee paralysis、数字锚定、节奏化交易及对 tokenomics 的误读。针对性的 harness 改造将捏造卖出规则比例由 57% 降至 3%，fee 主导观察由 32.5% 降至 10% 以下，资本部署率由 42.9% 提升至 78.0%。研究主张，对管理资本的 agents 应沿用户指令至 prompt、验证动作与结算的完整链路进行评估。
English abstract
We study reliability in autonomous language-model agents that translate user mandates into validated tool actions under real capital. The setting is DX Terminal Pro, a 21-day deployment in which 3,505 user-funded agents traded real ETH in a bounded onchain market. Users configured vaults through structured controls and natural-language strategies, but only agents could choose normal buy/sell trades. The system produced 7.5M agent invocations, roughly 300K onchain actions, about $20M in volume, more than 5,000 ETH deployed, roughly 70B inference tokens, and 99.9% settlement success for policy-valid submitted transactions. Long-running agents accumulated thousands of sequential decisions, including 6,000+ prompt-state-action cycles for continuously active agents, yielding a large-scale trace from user mandate to rendered prompt, reasoning, validation, portfolio state, and settlement. Reliability did not come from the base model alone; it emerged from the operating layer around the model: prompt compilation, typed controls, policy validation, execution guards, memory design, and trace-level observability. Pre-launch testing exposed failures that text-only benchmarks rarely measure, including fabricated trading rules, fee paralysis, numeric anchoring, cadence trading, and misread tokenomics. Targeted harness changes reduced fabricated sell rules from 57% to 3%, reduced fee-led observations from 32.5% to below 10%, and increased capital deployment from 42.9% to 78.0% in an affected test population. We show that capital-managing agents should be evaluated across the full path from user mandate to prompt, validated action, and settlement.
加载中…
点文件 → 加为 tab；按 Esc 关闭