# OpenClaw Claim-to-Evidence Map Example

## Project

- Project: OpenClaw
- Topic: embodied AI / robotics manipulation
- Target venue: TBD
- Scope date: 2026-03-10

## Major claims

| Claim ID | Claim | Status | Evidence IDs | Notes |
| --- | --- | --- | --- | --- |
| C1 | OpenClaw addresses manipulation tasks that require multi-stage perception, planning, and action. | partial | E01, E02 | Needs project artifact confirmation for exact scope and task definition. |
| C2 | OpenClaw can be positioned against recent embodied AI and robotics manipulation systems. | supported | E03, E04 | Requires verified citation set before related work writing. |
| C3 | OpenClaw improves long-horizon manipulation performance over strong baselines. | unsupported |  | Do not write this claim unless internal or external benchmark evidence is verified. |
| C4 | OpenClaw merits ablation analysis across perception, planning, and control components. | supported | E01, E05 | This is a paper-structure claim, not a result claim. |

## Evidence items

| Evidence ID | Classification | Source | What it supports | Limitations |
| --- | --- | --- | --- | --- |
| E01 | inference | Project method summary | Suggests a modular system design with multiple components. | Needs local code or design docs for verification. |
| E02 | unverified | Internal task notes | Possible manipulation task coverage. | Not yet tied to official task definitions or experiment artifacts. |
| E03 | verified_fact | Verified prior papers and official project pages | Existence of comparable prior work. | Exact comparison set depends on final scope. |
| E04 | verified_fact | Official benchmark and dataset pages | Available benchmarks, metrics, and task definitions. | Benchmark fit for OpenClaw still needs review. |
| E05 | verified_fact | Research paper conventions for system papers | Supports the need for ablations when multiple components contribute to performance. | Does not prove any OpenClaw result. |

## Gaps

- Missing verified internal result table
- Missing baseline comparison matrix
- Missing benchmark alignment decision
- Missing real-world or simulation-only positioning
- Missing contribution scope decision

## Safe writing notes

- C1 must stay scoped to system intent unless project artifacts confirm more
- C2 is safe for literature review planning once citations are verified
- C3 must not appear in any factual paper prose yet
- C4 can safely guide experiment planning and paper organization