DaoyuanLi2816 · Aug 29, 2026 · Aug 29, 2026 · Aug 29, 2026 · Aug 29, 2026 · Aug 29, 2026
diff --git a/CHANGELOG.md b/CHANGELOG.md
@@ -63,6 +63,20 @@ All notable changes to miniVERL are recorded here. The format follows
 - Kept PG-k1 on `hf_cached`: the BF16 external engine and NF4 actor agreed on
   greedy tokens in the qualification probe, but their sampled-token
   log-probabilities did not satisfy the preregistered numerical threshold.
+- Preserved the initial managed-server startup timing across later adapter
+  refreshes instead of overwriting it with no-op start checks.
+
+### Measured rollout evidence
+
+- Published the exact-wheel 24-cell RTX 4080 comparison for `hf_reference`,
+  `hf_cached` and managed vLLM across prompt lengths 128/512, response bounds
+  64/256/512, `n=1/4`, greedy and seeded sampling.
+- Selected vLLM 0.28.0 for the direct-GKD development path after a measured
+  1.65–4.55× advantage over `hf_cached` at 256/512 tokens, eight confirmed
+  policy refreshes, 11,820 MiB peak total GPU memory and clean teardown.
+- Recorded the blocking negative result: `hf_cached` did not reach the
+  preregistered 2× speedup over `hf_reference`. v0.11.0 remains unpublished
+  until the local backend is optimized and the frozen workload is rerun.
 
 ### Documentation narrative
 

diff --git a/PYPI.md b/PYPI.md
@@ -140,6 +140,21 @@ rewards and explicit group advantage composition; it has no task-quality claim.
 
 ## Measured systems evidence
 
+Rollout Runtime v2 now has a complete 24-cell RTX 4080 measurement across
+64/256/512-token responses, `n=1/4`, greedy and seeded sampling. Managed vLLM
+0.28.0 delivered 108.5–117.7 output tokens/s and was 1.65–4.55× faster than
+`hf_cached` in the 256/512-token cells, with 11.54 GiB peak total GPU memory,
+eight confirmed policy refreshes and clean teardown. It is the measured
+direct-GKD engine; PG-k1 stays on `hf_cached` because the engine log-probability
+probe exceeded its numerical threshold.
+
+The same evidence found the next optimization target: `hf_cached` reached
+about 1.8–1.9× on most seeded 256/512-token cells but did not meet the
+preregistered 2× gate, while the greedy path was usually slower than the legacy
+oracle. The development line therefore remains `0.11.0.dev0` rather than a
+published v0.11.0 release. See the [full runtime result](https://github.com/DaoyuanLi2816/mini-verl/blob/main/docs/benchmarks/rollout-runtime-v2.md)
+for every cell and artifact hash.
+
 The Qwen3-0.6B/1.7B developer workload consumed **32 distinct prompts**, used a
 64-token response bound, and completed **8 current-policy updates** at
 **3.1914 GiB peak reserved VRAM** on an RTX 4080. Median steady-state rollout,

diff --git a/README.md b/README.md
@@ -140,6 +140,21 @@ rewards and explicit group advantage composition; it has no task-quality claim.
 
 ## Measured systems evidence
 
+Rollout Runtime v2 now has a complete 24-cell RTX 4080 measurement across
+64/256/512-token responses, `n=1/4`, greedy and seeded sampling. Managed vLLM
+0.28.0 delivered 108.5–117.7 output tokens/s and was 1.65–4.55× faster than
+`hf_cached` in the 256/512-token cells, with 11.54 GiB peak total GPU memory,
+eight confirmed policy refreshes and clean teardown. It is the measured
+direct-GKD engine; PG-k1 stays on `hf_cached` because the engine log-probability
+probe exceeded its numerical threshold.
+
+The same evidence found the next optimization target: `hf_cached` reached
+about 1.8–1.9× on most seeded 256/512-token cells but did not meet the
+preregistered 2× gate, while the greedy path was usually slower than the legacy
+oracle. The development line therefore remains `0.11.0.dev0` rather than a
+published v0.11.0 release. See the [full runtime result](docs/benchmarks/rollout-runtime-v2.md)
+for every cell and artifact hash.
+
 The Qwen3-0.6B/1.7B developer workload consumed **32 distinct prompts**, used a
 64-token response bound, and completed **8 current-policy updates** at
 **3.1914 GiB peak reserved VRAM** on an RTX 4080. Median steady-state rollout,

diff --git a/README.zh-CN.md b/README.zh-CN.md
@@ -129,6 +129,18 @@ group advantage 组合；目前不对任务质量作结论。
 
 ## 实测系统证据
 
+Rollout Runtime v2 已在 RTX 4080 上完成 24-cell 测量，覆盖 64/256/512-token
+response、`n=1/4`、greedy 与 seeded sampling。Managed vLLM 0.28.0 达到
+108.5–117.7 output tokens/s；在 256/512-token cell 中比 `hf_cached` 快
+1.65–4.55×，peak total GPU memory 为 11.54 GiB，并通过 8 次 policy refresh 与
+完整 teardown。它是实测的 direct-GKD engine；engine log-probability probe 超出数值
+门槛，因此 PG-k1 继续使用 `hf_cached`。
+
+同一份证据也给出了下一步优化目标：`hf_cached` 在多数 seeded 256/512-token cell 中
+约为旧路径的 1.8–1.9×，没有达到预注册的 2× gate；greedy 路径通常比旧 oracle 更慢。
+因此开发线保持 `0.11.0.dev0`，尚未发布 v0.11.0。每个 cell 与 artifact hash 见
+[完整 runtime 结果](docs/benchmarks/rollout-runtime-v2.md)。
+
 Qwen3-0.6B/1.7B developer workload 消费 **32 个不同 prompt**，response 上限为 64 token，
 在 RTX 4080 上完成 **8 次 current-policy update**，**peak reserved VRAM 为 3.1914 GiB**。
 稳态 rollout、teacher scoring 与 update 的中位耗时分别为 9.7200、0.4864 与 2.3260 秒。

diff --git a/benchmarks/README.md b/benchmarks/README.md
@@ -80,6 +80,7 @@ Listed so you can check that this repository does not.
 
 | File | Hardware | What it shows |
 | --- | --- | --- |
+| `results/rollout-runtime-v2-rtx4080.json` | WSL2 RTX 4080 16 GB | Exact-wheel 24-cell `hf_reference` / `hf_cached` / vLLM comparison. vLLM passed the direct-GKD engine gates; `hf_cached` missed the preregistered 2× speed gate, so v0.11.0 remains blocked. [Result, method and hashes.](../docs/benchmarks/rollout-runtime-v2.md) |
 | `results/rollout-runtime-v2-hf-reference.json` | WSL2 RTX 4080 16 GB | Pre-v0.11 actor-rollout systems baseline: all 24 preregistered prompt/response/n/sampling cells completed. It freezes throughput and memory for the future `hf_cached` comparison; teacher scoring, updates, full-cycle timing and typed grouped semantics are not measured here. [Method, table and boundaries.](../docs/benchmarks/rollout-runtime-v2.md) |
 | `results/alignment-external-v1.json` | RTX 4080 16 GB | **Preregistered early stop, not a method comparison.** Two declared lineages and eight candidates all measured 0/64 retained JSONNav utility against an unchanged 20% floor. No checkpoint, teacher, continuation arm or final-test result exists. [Study, correction provenance and evaluator boundary.](../docs/alignment-external/alignment-external-v1.md) |
 | `results/alignment-lab-v1.json` | RTX 4080 16 GB | Alignment Lab v1: six methods from one Qwen3-0.6B SFT checkpoint, three seeds and 48 paired policy tasks per arm. The SFT start saturated alignment and utility; no continuation method improved it, so the pilot recommends no online teacher querying. [Report, figures and caveats.](../docs/alignment-lab/alignment-lab-v1.md) |