perf(agentx): refresh dsv4-fp4-gb300-dynamo-vllm-agentic-mtp-disagg harness

[cquil11]

Summary

• Add a perf changelog entry for dsv4-fp4-gb300-dynamo-vllm-agentic-mtp-disagg.
• Refresh the submission with the up-to-date AgentX harness.

Validation

• Parsed perf-changelog.yaml successfully.
• git diff --check passes.

[github-actions]

Thanks for the contribution! Please reach out to respective companies' CODEOWNER to fill in the latest PR_REVIEW_CHECKLIST.md before pinging core maintainer on Slack for review. In order for the signoff PR check bot to trigger, you must follow the PR_REVIEW_CHECKLIST.md template correctly, including the phrase As a PR reviewer and CODEOWNER, I have reviewed this and have.

For PR verification, add the full-sweep-fail-fast label (strongly recommended) to this PR — the benchmark sweep only runs on labeled PRs. Use full-sweep-enabled only if you need matrix jobs to keep running past a failure.

PR authors are responsible for ensuring that after merging, all GitHub Action jobs fully pass. A lot of the time, failures are just flakes and simply re-running the failed jobs will fix it. See GitHub's docs on re-running failed jobs

---

感谢你的贡献！请联系相应公司的 CODEOWNER 填写最新的 PR_REVIEW_CHECKLIST.md，然后再在 Slack 上联系核心维护者进行审阅。为了触发 signoff PR 检查机器人，你必须正确遵循 PR_REVIEW_CHECKLIST.md 模板，包括保留英文语句 As a PR reviewer and CODEOWNER, I have reviewed this and have。

如需进行 PR 验证，请为此 PR 添加 full-sweep-fail-fast 标签（强烈推荐）— 基准测试 sweep 仅在带有标签的 PR 上运行。仅当需要矩阵任务在失败后继续运行时才使用 full-sweep-enabled。

PR 作者有责任确保合并后所有 GitHub Action 任务完全通过。 很多时候失败只是偶发抖动（flake），重新运行失败的任务即可解决。参见 GitHub 关于重新运行失败任务的文档

[claude]

Claude finished @cquil11's task in 3m 7s —— View job

---

Review of PR #2519

• Gather context and read the perf-changelog.yaml diff
• Verify new entry is appended to the end of the file and is valid
• Post review feedback

LGTM - no blocking issues found

• The new entry is appended to the very end of perf-changelog.yaml (after the previous last entry, PR perf(agentx): refresh Qwen3.5 GB300 recipes with correct metrics #2477), preserving chronological order.
• Entry format matches existing entries (config-keys, scenario-type, description, pr-link), and the referenced key dsv4-fp4-gb300-dynamo-vllm-agentic-mtp-disagg exists in configs/nvidia-master.yaml.
• git diff --check origin/main...HEAD passes; no master config files were modified, so no additional changelog coupling checks apply.

[claude]

LGTM — trivial changelog-only addition.

Extended reasoning...

Overview

The PR appends a single new entry to perf-changelog.yaml (config-keys, scenario-type, description, pr-link) for the dsv4-fp4-gb300-dynamo-vllm-agentic-mtp-disagg recipe. No other files are touched, and the referenced config key already exists in configs/nvidia-master.yaml from prior commits.

Security risks

None. This is a static YAML documentation/changelog file with no executable logic, no user input handling, and no auth/crypto/permissions surface.

Level of scrutiny

Minimal scrutiny is appropriate — this is a pure metadata/changelog append, structurally identical to dozens of prior entries in the same file, and it parses cleanly (YAML list append at end of file).

Other factors

The bug hunting system found no bugs. One candidate issue was raised and ruled out — that the changelog claims a 'harness refresh' without an accompanying config/harness diff in this PR. Checking git history, this matches an established pattern in this repo where changelog entries documenting a harness refresh are sometimes recorded as their own commit separate from the underlying config change, so this isn't anomalous for this codebase.

[github-actions]

see unofficial run visualizer at https://inferencex.semianalysis.com/inference?unofficialRun=31150508200
see unofficial run visualizer at https://inferencex.semianalysis.com/evaluation?unofficialRun=31150508200

[github-actions]

see unofficial run visualizer at https://inferencex.semianalysis.com/inference?unofficialRun=31150508200
see unofficial run visualizer at https://inferencex.semianalysis.com/evaluation?unofficialRun=31150508200

[github-actions]

see unofficial run visualizer at https://inferencex.semianalysis.com/inference?unofficialRun=31152069277
see unofficial run visualizer at https://inferencex.semianalysis.com/evaluation?unofficialRun=31152069277

[github-actions]

see unofficial run visualizer at https://inferencex.semianalysis.com/inference?unofficialRun=31153318004
see unofficial run visualizer at https://inferencex.semianalysis.com/evaluation?unofficialRun=31153318004

[github-actions]

see unofficial run visualizer at https://inferencex.semianalysis.com/inference?unofficialRun=31153942559
see unofficial run visualizer at https://inferencex.semianalysis.com/evaluation?unofficialRun=31153942559