From 6c5b4bb0f9376546bfa8b55c678b6256879ad510 Mon Sep 17 00:00:00 2001
From: Gotham-Zolio <18781106300@163.com>
Date: Fri, 11 Sep 2026 14:44:34 -0400
Subject: [PATCH] docs: replace "within a few minutes" with the curve E6
 measured

The quickstart said the return climbs out of the -300s within a few minutes.
It is true in the weakest sense and useless in every other: the curve is
noisy enough early that a seed is still dipping into the -300s at step 20k,
and someone who watches for three minutes and sees noise concludes the thing
does not learn.

Now it says what the three seeds did - across zero at about 60k steps,
1928 +/- 224 at 500k, about a hundred minutes on the CPU-only machine that
measured it - tells the reader the first few minutes are noise, and links
the run. Both languages.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
---
 docs/index.md    | 14 ++++++++++----
 docs/index.zh.md | 10 +++++++---
 2 files changed, 17 insertions(+), 7 deletions(-)

diff --git a/docs/index.md b/docs/index.md
index 3841d05..e5f0c83 100644
--- a/docs/index.md
+++ b/docs/index.md
@@ -22,10 +22,16 @@ plugrl-run-env-client mujoco-v1 \
     --num-envs 1 --num-episodes 600 --runner.replan-steps 1 --runner.seed 0
 ```
 
-Episode return climbs out of the -300s within a few minutes. `HalfCheetah-v5`
-has a 17-dimensional observation and a 6-dimensional action, which are
-exactly `fpo-policy`'s defaults, so nothing needs configuring. The
-environment needs `plugrl-env-client[mujoco]`.
+`HalfCheetah-v5` has a 17-dimensional observation and a 6-dimensional
+action, which are exactly `fpo-policy`'s defaults, so nothing needs
+configuring. The environment needs `plugrl-env-client[mujoco]`.
+
+Episode return starts near -300. Across three seeds it is still dipping back
+into the -300s at step 20k, the mean crosses zero at about 60k, and by 500k
+steps it reaches **1928 ± 224** — roughly a hundred minutes on the CPU-only
+machine that measured it. The first few minutes are noise, so judge it over
+tens of thousands of steps rather than the first screenful. Curve, seeds and
+logs: [E6](https://github.com/PlugRL/plugrl-server/tree/main/experiments/e6-first-learning-curve).
 
 !!! warning "`--algo.buffer-size` is not decoration"
 
diff --git a/docs/index.zh.md b/docs/index.zh.md
index 4661db3..4ed2462 100644
--- a/docs/index.zh.md
+++ b/docs/index.zh.md
@@ -21,9 +21,13 @@ plugrl-run-env-client mujoco-v1 \
     --num-envs 1 --num-episodes 600 --runner.replan-steps 1 --runner.seed 0
 ```
 
-几分钟内 episode 回报就会从 -300 附近爬上来。`HalfCheetah-v5` 的观测是 17 维、
-动作是 6 维，**正好是 `fpo-policy` 的默认值**，所以不需要任何配置。
-环境端需要 `plugrl-env-client[mujoco]`。
+`HalfCheetah-v5` 的观测是 17 维、动作是 6 维，**正好是 `fpo-policy` 的默认值**，
+所以不需要任何配置。环境端需要 `plugrl-env-client[mujoco]`。
+
+episode 回报从 -300 附近起步。三个随机种子里，到第 2 万步仍有种子会掉回 -300 区间，
+均值在约 6 万步处越过零点，到 50 万步达到 **1928 ± 224** —— 在做这次测量的纯 CPU
+机器上大约一百分钟。**最初几分钟全是噪声**，要按几万步的尺度看，而不是看屏幕上
+最先出现的那几行。曲线、种子与日志：[E6](https://github.com/PlugRL/plugrl-server/tree/main/experiments/e6-first-learning-curve)。
 
 !!! warning "`--algo.buffer-size` 不是装饰"