Add HumanoidArena sessions without expert action demonstrations; link arXiv and add citation

README.md

@@ -1,15 +1,18 @@
 ---
 license: cc-by-4.0
-pretty_name: RAX robot session logs (Franka, RB-Y1)
-tags: [robotics, llm-agent, claude-code, franka, rainbow-rby1, manipulation, mobile-manipulation]
+pretty_name: RAX robot session logs (Franka, RB-Y1, HumanoidArena)
+tags: [robotics, llm-agent, claude-code, franka, rainbow-rby1, manipulation, mobile-manipulation, humanoid, simulation]
 ---
 
 # RAX robot session logs
 
-Complete logs of Claude Code agents learning manipulation tasks on two real robots through RAX
-([paper](https://robot-agent-exploration.com/rax.pdf), [project page](https://robot-agent-exploration.com/),
-[code](https://github.com/robot-agent-exploration/rax),
-[skill library](https://github.com/robot-agent-exploration/skills-lib)). Each session holds:
+- arXiv: https://arxiv.org/abs/2610.22496
+- Project page: https://robot-agent-exploration.com/
+- Code: https://github.com/robot-agent-exploration/rax
+- Skill library: https://github.com/robot-agent-exploration/skills-lib
+
+Complete logs of Claude Code agents learning manipulation tasks on two real robots and a simulated humanoid through
+RAX. Each session holds:
 - the agent's full transcripts, including subagents and tool results;
 - every robot tool call with its arguments, result and error;
 - the robot state log and camera recordings;
@@ -22,6 +25,9 @@ Complete logs of Claude Code agents learning manipulation tasks on two real robo
 - **RB-Y1** (2026-10-08 to 2026-10-10): 7 sessions on a Rainbow Robotics RB-Y1 wheeled bimanual manipulator with
   a head RGB-D camera: move a box and a bucket onto a cabinet, tuck an office chair, stack cups, and a compositional
   task that chains them.
+- **HumanoidArena** (2026-10-12 to 2026-10-13): 3 sessions on a simulated Unitree G1 humanoid in HumanoidArena,
+  without expert action demonstrations: score a goal, pick and place a box, and move a hammer to a basket on another
+  desk. Their layout is described in [`humanoid-arena/README.md`](humanoid-arena/README.md).
 
 ## Layout
 
@@ -38,9 +44,11 @@ rby1/extras/skill-library-commit-ids.tsv   old -> new commit ids of the rebuilt
 rby1/extras/robot-models.zip       the vendor URDFs and meshes the sessions used
 rby1/extras/characterization.zip   hardware test results from bring-up
 rby1/extras/phone-videos/rby1-{1,2,3}.mp4   phone videos filming the compositional session (no audio)
+humanoid-arena/README.md           layout of the HumanoidArena sessions
+humanoid-arena/<task>/<task>.tar   one HumanoidArena session (football, pickplace_box, doubledesk)
 ```
 
-Inside a session tar:
+Inside a Franka or RB-Y1 session tar (the HumanoidArena tars are described in `humanoid-arena/README.md`):
 
 | Path | Contents |
 |---|---|
@@ -103,6 +111,14 @@ a task:
 - `20261009T002555Z_controller_check_fae961`: operator controller characterization, no agent
 - `20261009T232457Z_stack_bowl_and_place_basket_bad3d4`: 1-minute start, superseded by the compositional session
 
+### HumanoidArena
+
+| Session | Task | Exploration budget | Formal success |
+|---|---|---|---|
+| `humanoid-arena/football` | Football: score a goal | 64 | 59 / 60 |
+| `humanoid-arena/pickplace_box` | P&PBox: pick and place a box | 128 | 42 / 60 |
+| `humanoid-arena/doubledesk` | DoubleDesk: move a hammer to a basket on the other desk | 256 | 28 / 60 |
+
 ## Privacy and redactions
 
 - **Faces are blurred.** Faces were detected with YuNet at full resolution in:
@@ -129,7 +145,8 @@ a task:
   colour of the recording.
 - **Paths.** Lab paths and user names are rewritten. The robot runtime is `/opt/rax/runtime`, the harness checkout
   `/opt/rax`, the host home `/home/robot`. Other projects that appear in the agent's sandbox deny-list are shown as
-  `other-project-N` / `other-data-N`.
+  `other-project-N` / `other-data-N`. In the HumanoidArena transcripts, host storage paths and user and group
+  ids are shown as `<host-storage>`, `<uid>` and `<gid>`.
 - **Accounts.** Agent account emails are redacted. The agent CLI's local state is not included: account file,
   backups, plugin marketplace copies, caches, shell snapshots, job state, task lists and file-edit backups. No API
   keys or tokens were present.
@@ -150,3 +167,14 @@ a task:
 - To replay a transcript with Claude Code, place the session at `/opt/rax/runtime/sessions/<session>` and run
   `CLAUDE_CONFIG_DIR=<session>/claude-home claude --resume` in `<session>/agent-workspace`. The robot tools need
   the original robot and harness.
+
+## Citation
+
+```bibtex
+@article{huang2026scaling,
+  title={Scaling Online Exploration for Robot Agents},
+  author={Huang, Runhan and Zhang, Xiangcheng and Chen, Haoyu and Li, Yitong and Wang, Yen-Jen and Lyu, Bohan and Ye, Haotian and Hong, Chuye and Wang, Yixuan and Su, Chloe and others},
+  journal={arXiv preprint arXiv:2610.22496},
+  year={2026}
+}
+```

humanoid-arena/README.md

@@ -0,0 +1,72 @@
+# RAX on HumanoidArena without expert action demonstrations
+
+Run records of RAX on three HumanoidArena tasks without expert action demonstrations.
+
+| Directory | Task | Exploration budget | Formal success |
+|---|---|---|---|
+| `football/` | Football | 64 | 59 / 60 |
+| `pickplace_box/` | P&PBox | 128 | 42 / 60 |
+| `doubledesk/` | DoubleDesk | 256 | 28 / 60 |
+
+## Where to find what
+
+### Exploration logs
+
+| What | Where |
+|---|---|
+| Agent transcripts | `agent/transcripts/main.jsonl` is the main agent's conversation, one message per line, including every tool call it issued and the result it got back. `agent/transcripts/subagent-<id>.jsonl` are the subagents' conversations in the same format. |
+| Tool calls to the robot | `episodes/<run>/mcp.jsonl` logs every robot tool call (`humanoid_step`, `humanoid_observe`, ...) in that run as request/result pairs. The full response of each call and the camera images it returned are in `episodes/<run>/calls/<n>/`. |
+| Development outcomes | Each exploration episode is an `episodes/interactive-<id>/` directory with its own `result.json`. `results/exploration_and_rehearsals.json` lists the outcome of every exploration episode in one file. |
+| Rehearsal outcomes | Each rehearsal of a candidate controller is an `episodes/rehearsal-<id>/` directory containing the rehearsed code (`code/`), its logs (`stdout.log`, `stderr.log`, `logs.json`) and `result.json`. These outcomes are also listed in `results/exploration_and_rehearsals.json`. |
+| Memory notes | `agent/workspace/memory/` holds the notes the agent kept during the run (`PROGRESS.md`, `TIPS.md` and, where present, `INDEX.md` and topic notes). |
+
+### Final controllers
+
+| What | Where |
+|---|---|
+| Submitted code bundle | `submission/` is the exact bundle the agent submitted at the end of exploration. `submission/code/<name>/` is the controller code that was run in the formal evaluation, together with its supporting assets (e.g. `robot.json`, `arm_ref.npy`). `submission/memory/` is the memory frozen with the bundle, and `submission/package.json` its metadata. |
+| Formal evaluation | `results/formal.json` gives the submitted bundle id, the success count and the outcome of each of the 60 formal episodes. |
+
+The controller directory names are the agent's own choices: `football`, `pickplace`
+and `hammer` (for DoubleDesk).
+
+## Layout
+
+Each task is packed as `<task>/<task>.tar`. Extracted, every task directory has the same layout:
+
+```
+<task>/
+├── setting.json                  # experiment setting (task, observation, budgets, resource limits, model)
+├── agent/
+│   ├── workspace/                # the agent's workspace at the end of the run
+│   │   ├── TASK.md               # task description given to the agent
+│   │   ├── AGENTS.md, START_PROMPT.md, MCP_SESSION.md, MCP_CONTRACT.json
+│   │   ├── .agents/              # skills available to the agent
+│   │   ├── code/                 # controller code written by the agent
+│   │   ├── memory/               # notes the agent kept
+│   │   └── experiments/          # the agent's experiment scripts and outputs
+│   └── transcripts/
+│       ├── main.jsonl            # main agent conversation (messages and tool calls)
+│       └── subagent-<id>.jsonl   # subagent conversations
+├── submission/                   # final controller submitted for formal evaluation
+│   ├── code/<name>/
+│   ├── memory/
+│   └── package.json
+├── episodes/
+│   ├── interactive-<id>/         # one exploration episode
+│   │   ├── mcp.jsonl             # robot tool-call log
+│   │   ├── calls/<n>/            # response.json and images returned by each call
+│   │   ├── frames/               # recorded frame sequences
+│   │   └── result.json
+│   └── rehearsal-<id>/           # one rehearsal of a candidate controller
+│       ├── code/                 # the controller that was rehearsed
+│       ├── calls/, frames/, exports/, mcp.jsonl
+│       ├── stdout.log, stderr.log, logs.json
+│       └── result.json
+└── results/
+    ├── formal.json               # formal evaluation of the submitted controller
+    └── exploration_and_rehearsals.json   # outcome of every exploration episode and rehearsal
+```
+
+Host-specific paths and IDs in the transcripts are replaced with `<host-storage>`,
+`<uid>` and `<gid>`.

humanoid-arena/doubledesk/doubledesk.tar

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1407c8f5e89129713f35b1b8d28b001d57f387a77b00dfcf316e21b50c336f27
+size 30692710400

humanoid-arena/football/football.tar

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:dd73afdfaad71a403e15ea5eea3abfca3900b2d727711bf23174a2967bdb20e9
+size 11373137920

humanoid-arena/pickplace_box/pickplace_box.tar

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1b56182fd98cabdd608c32fea630206fcfa12e4cad2d916eabbdc66e262c9839
+size 8773693440