Skip to content

#Reviews/benchmarks

0 items today
8/22Sat
8/20Thu
8/19Wed
  1. Cline · Blog71

    Cline's Evaluation Methodology and Trace for Open-Weight Models

    Cline has open-sourced its evaluation methodology for open-weight models, along with a hill-climbing score and trace worth over one thousand dollars, available for download and analysis. The post lays out five heuristics from the Hill Climber's Checklist: set a North Star metric, quantify noise, break down failure modes by task/model/vendor, don't assume more thinking is always better, and keep a private evaluation set.

    Why it matters: Cline shares its evaluation methodology and a trace worth over a thousand dollars, offering five transferable hill-climbing heuristics that teams building their own harness can reference.

8/17Mon
8/10Mon
  1. Martin Fowler · Exploring Generative AI74

    智能体循环里的 TDD 是形式还是真价值?Martin Fowler 的对比实验

    Martin Fowler 用 Sonnet 4.6 生成、Opus 4.8 盲评的方式,对小型、中型和较大型三类业务逻辑任务分别跑 TDD 与非 TDD 方案,结论是两者质量没有明显差异,非 TDD 方案在设计和测试质量上还多次略高,变异分数也没有实质差别。

    Awaiting translation

    Why it matters: 作者用同一批任务对比 TDD 与非 TDD 智能体实现,给出 token 成本与设计质量差异,并反思哪些 TDD 收益在智能体循环里失效。

8/5Wed
  1. Chen Dahuang · AI Coding 实录66

    DeepSeek V4 Flash 正式版深度体验:便宜、快、1M 上下文、内置搜索

    作者深度体验几天 DeepSeek V4 Flash 0731 正式版后总结:便宜到跑批处理、Agent 循环和几十轮对话账单基本无感,速度快到配合 Agent 工具循环每步几秒内完成,1M 上下文可容纳整个仓库和完整对话历史、无需频繁 compact,结合 Cache 打折长上下文成本还能再降。

    Awaiting translation

7/30Thu
  1. Terminal-Bench · News60

    Terminal-Bench 3.0 is out: 74 tasks across 7 domains, with the strongest model passing about 34%

    The Terminal-Bench team releases Terminal-Bench 3.0, whose first version spans 7 domains and 74 tasks, with the strongest model passing about 34%. Building on Terminal-Bench 2.1, this release broadens task diversity and adds CI/CD, semantic versioning, and result migration to keep improving the benchmark.

    Why it matters: Terminal-Bench 3.0 rebuilds the benchmark with 74 tasks and CI/CD-based versioning, so readers can see how the new benchmark separates models.

  2. Terminal-Bench · News62

    Terminal-Bench ships new Harbor features, turning the benchmark into a versioned asset that keeps getting updated

    The Terminal-Bench team has shipped a batch of new Harbor features that let datasets be released by version and let leaderboards migrate to new versions by reusing, re-evaluating, or rerunning trials. Tasks use semantic versioning: patch-level changes reuse old results as-is, validator changes only require re-evaluating saved artifacts, and only major changes that significantly alter the agent environment require a rerun. Dataset versions follow the highest version number among the tasks, and leaderboards use diffs to rerun only the tasks with major changes.

    Why it matters: The Terminal-Bench team maintains the benchmark like software, laying out concrete mechanisms for task semantic versioning and leaderboard upgrades that you can carry over to your own evaluation pipeline.

7/25Sat
  1. Cline · Blog79

    Cline 用递归自我改进让智能体自主把 Kimi K3 在 Terminal-Bench 2.1 提到 88.8%

    Cline 用一条提示词启动 17 小时连续运行的编码智能体,以 GPT-5.6-Sol 为 leader 模型,把 Kimi K3 在 Terminal-Bench 2.1 上的成绩从基线 69/89(77.5%,$79)提升到确认运行 79/89(88.8%,$49.8),超过 Moonshot 自报的 88.3%。

    Awaiting translation

    Why it matters: Cline 用一条提示词让智能体自主完成 17 小时爬山,把 Terminal-Bench 2.1 从 77.5% 提到 88.8%,可看具体修了哪些 harness 问题。

7/12Sun
6/18Thu
  1. Terminal-Bench · News60

    Terminal-Bench 发布 Challenges 长周期智能体基准

    Terminal-Bench 发布 Challenges,一种长周期、高 token 消耗的单任务基准,要求智能体在无时间与资源限制下自主完成整个项目,首批开放三个挑战。

    Awaiting translation

    Why it matters: Terminal-Bench 官方推出长周期单任务基准,并公开三个挑战的实测失败模式,可供评估智能体长时自主能力时参考。

6/15Mon
6/8Mon
6/6Sat
6/1Mon
4/30Thu
  1. Augment Code · Blog71

    Augment Code put Karpathy-style rules to the test: the coding agent didn’t write better code, but it was cheaper and faster

    In AGENTS.md, Augment Code front-loads roughly 2.5k characters of Karpathy-style coding rules, then runs 40 OpenClaw PRs through Auggie, Claude Code, and Codex for comparison.

    Why it matters: A head-to-head test of three coding agents on the same set of PRs shows that prompt constraints mainly cut costs rather than improve quality, and it also surfaces differences between the harnesses.

4/24Fri
  1. Lovable · Blog38

    Lovable 早期测试 GPT-5.5:最难任务通过率 41.6%,比 GPT-5.4 提升 12.5%

    Lovable 在早期访问中测试 GPT-5.5,其内部基准显示最难任务通过率从 GPT-5.4 的 36.9% 升至 41.6%,每次请求工具调用减少 23.1%,用户卡住的消息占比下降 9.9%。GPT-5.5 每请求输出 token 减少 33%,日常任务成本效率提升约 15%,将很快向 Lovable 构建者开放。

    Awaiting translation

4/22Wed
  1. Augment Code · Blog88

    Augment Code Tests AGENTS.md: A Good File Is Like a Model Upgrade, a Bad One Is Worse Than Nothing

    Augment Code pulled dozens of AGENTS.md files from its own monorepo and used its internal benchmark suite AuggieBench to compare how the same tasks performed with and without the file. The best files delivered a quality boost equivalent to upgrading from Haiku to Opus, while the worst made the output worse than having no AGENTS.md at all.

    Why it matters: Augment Code used internal benchmarks to quantify how much the different ways of writing AGENTS.md actually differ, so readers can adjust their own repo's documentation structure accordingly.

4/9Thu
4/7Tue
3/20Fri
  1. Terminal-Bench · News34

    Terminal-Bench 如何设计好的基准任务

    Terminal-Bench 发布任务设计指南,提出好的基准任务应具备对抗性、难度和可读性:指令像给资深工程师那样清晰直接,测试只验证结果而非实现细节,允许替代解法并防止 reward hacking。难度应来自问题本身,而非苛刻的输出格式或隐藏假设。作者建议亲自运行任务、检查容器、执行 oracle 并观察真实智能体轨迹,失败运行尤其能揭示任务是否真正困难。

    Awaiting translation

3/12Thu
3/8Sun
3/5Thu
  1. Terminal-Bench · News30

    Terminal-Bench 3.0 公开征集任务贡献

    Terminal-Bench 3.0 已进入开发阶段,目标收录 100 个多样化任务,发布时最强模型的解决率不超过 30%。任务需为可通过命令行完成并程序化验证的真实计算机工作,覆盖更长周期、多微服务/文件系统/数据库等更丰富环境及专家级知识,合并窗口开放至 5 月底。贡献者提交一个被接受的任务即可在最终版本中获得署名。

    Awaiting translation

3/4Wed
2/23Mon
2/5Thu
  1. Hacker News · AI Code Review 讨论42

    Qodo 发布 AI 代码审查基准 1.0:100 个 PR、580 个问题

    Qodo 研究团队发布 AI 代码审查基准 1.0,向真实已合并 PR 注入缺陷,用 100 个 PR、580 个问题同时评测代码正确性与最佳实践合规性。在对比 Qodo 模型与 7 家主流 AI 代码审查平台的评测中,Qodo 以 60.1% 的 F1 分数领先,基准及评测结果已在 GitHub 公开。

    Awaiting translation

1/19Mon
  1. Kondasamy Jayaraman · Engineering Blog60

    RAG 入门:构建可用检索系统时没人告诉你的那些事

    作者结合自己搭建多个 RAG 系统的经验,把 RAG 拆成检索与生成两步,指出检索质量决定答案质量,提示词工程无法弥补糟糕的上下文。文中给出具体取舍:分块 300-500 token 并保留 20-30% 重叠,纯向量检索只能找到约 75% 的相关文档,混合检索可提升到 87%,重排能带来 20-35% 的准确率提升但增加 200-500ms 延迟。

    Awaiting translation

1/8Thu
11/30Sun
11/7Fri
  1. Terminal-Bench · News62

    Terminal-Bench ships version 2.0 and an optimized Harbor evaluation package

    Terminal-Bench has released version 2.0 and the Harbor package. The former is a more rigorously validated, harder benchmark for evaluating agents; the latter is for evaluating and optimizing agents. Harbor rewrites Terminal-Bench's test harness, supports deploying containers in the cloud, provides rollout interfaces for RL and SFT, and works with any agent you can put in a container.

    Why it matters: Terminal-Bench 2.0 and Harbor are released together, so readers can see how the agent evaluation benchmark is validated and how to scale it in the cloud.

10/15Wed
  1. Martin Fowler · Exploring Generative AI74

    拆解 Spec-Driven Development:Kiro、spec-kit 与 Tessl 三种工具实测

    Martin Fowler 试用 Kiro、spec-kit 和 Tessl 三款自称实现 spec-driven development(SDD)的工具,把 SDD 归纳为 spec-first、spec-anchored、spec-as-source 三个层次,并指出目前所有方案都停留在 spec-first。

    Awaiting translation

    Why it matters: 作者亲手试用 Kiro、spec-kit 和 Tessl 三款 SDD 工具,给出 spec-first、spec-anchored、spec-as-source 三层划分,并指出小任务被过度规格化的问题。

9/9Tue
  1. Terminal-Bench · News38

    Terminal-Bench 排行榜回应超时争议:OB-1 重新提交成绩后重回榜首

    OpenBlock Labs 的智能体 OB-1 在按正确超时限制重新提交成绩后,重新成为 Terminal-Bench 排行榜得分最高的智能体。此前该提交将数据集中每个任务的超时统一改为固定 30 分钟,而多数任务原本限时 5 分钟,导致 75 个任务超时被放宽、2 个不变、3 个被缩短。

    Awaiting translation

7/15Tue
6/25Wed
6/20Fri
5/23Fri
  1. Terminal-Bench · News32

    Anthropic 在 Claude 4 模型卡中引入 Terminal-Bench,Claude 4 Opus 创下 43.2% 新 SOTA

    Anthropic 将 Terminal-Bench 列为 Claude 4 模型卡七项基准之一,Claude 4 Opus 在 Terminal-Bench-Core 上取得 43.2% 的 SOTA 成绩。Dario Amodei 在 Code with Claude 主题演讲中也提及该基准。Terminal-Bench 团队表示将在未来几天验证 Claude 4 的表现并更新官方排行榜。

    Awaiting translation

5/19Mon
  1. Terminal-Bench · News48

    Terminal-Bench 推出研究预览版智能体 Terminus

    Terminal-Bench 团队发布研究预览版智能体 Terminus,用于在终端中一致地评估语言模型驱动自主智能体的能力,发布时其性能在 Terminal-Bench 上仅次于 Claude Code。Terminus 采用单工具设计,仅通过 tmux 会话发送标准按键操作,并借助 LiteLLM 支持几乎所有 API 或本地托管模型,且完全自主运行、不请求用户输入。

    Awaiting translation