---
title: LikePhys
description: "Oxford likelihood probe: does a video diffusion model prefer physically valid clips over matched implausible ones? No generated-video input."
---

LikePhys (University of Oxford, ICLR 2026) is a **training-free likelihood probe**, not a generation bench. It does **not** score videos you generate. For each of **12 simulated scenarios** (rigid bodies, fluids, deformables, optics), it pairs a physically valid clip with matched implausible variations and asks whether the diffusion model assigns **higher likelihood** (lower denoising ELBO) to the valid clip. The primary metric is **mis-rank rate** — how often the model prefers the implausible variation. **Lower is better** everywhere.

That is the opposite job from [Physics-IQ](/docs/evaluation/benchmark-hub/physics-iq) (continue a filmed scene) and from [PhyGenBench](/docs/evaluation/benchmark-hub/phygenbench) (judge a generated T2V clip).

Paper: [LikePhys](https://arxiv.org/abs/2510.11512). Project: [yuanjianhao508.github.io/LikePhys](https://yuanjianhao508.github.io/LikePhys/). Code: [YuanJianhao508/LikePhys](https://github.com/YuanJianhao508/LikePhys). Clips: [JianhaoDYDY/LikePhys-Benchmark](https://huggingface.co/datasets/JianhaoDYDY/LikePhys-Benchmark) (920 clips).

WorldFoundry **recomputes the official mis-rank formulas** in tree (`compute_misrank_normalized` + `read_exp_final.py` filters). Driving the upstream ELBO probe needs a caller-supplied LikePhys checkout (`WORLDFOUNDRY_LIKEPHYS_EVALUATOR_ROOT`), CUDA, and the probed checkpoints — WorldFoundry does not vendor the forked diffusers pipelines.

<BenchmarkPageSections id="likephys" locale="en" />

## How to run

Import `results_<model>.json` files (or a directory of `<scenario>/results_<probe_model>.json`) and recompute mis-rank:

```bash
worldfoundry-eval zoo benchmark-run \
  --benchmark-id likephys \
  --mode official-validation \
  --official-results-path <likephys_results_root> \
  --output-dir tmp/likephys/official-validation \
  --json
```

Use `--probe-model` when a results root holds several backends. `--no-report-filter` keeps variations the paper's `filter_config` excludes.

Run the upstream ELBO probe for one model on one scenario (CUDA + LikePhys checkout):

```bash
hf download JianhaoDYDY/LikePhys-Benchmark --repo-type dataset --local-dir "${WORLDFOUNDRY_LIKEPHYS_DATA_ROOT}"

WORLDFOUNDRY_LIKEPHYS_EVALUATOR_ROOT=/path/to/LikePhys \
${WORLDFOUNDRY_UNIFIED_PYTHON} \
  worldfoundry/evaluation/tasks/execution/runners/likephys/run_likephys_official_runner.py \
  --run-official \
  --probe-model wan2.1-T2V-1.3b \
  --scenario ball_drop \
  --benchmark-data-root "${WORLDFOUNDRY_LIKEPHYS_DATA_ROOT}" \
  --output-dir tmp/likephys/probe \
  --json
```

`--dry-run` prepares the workspace and prints commands. Outputs include `scorecard.json` (primary `likephys_misrank_rate`) and `misrank_breakdown.json`.

[← Benchmark Hub](/docs/evaluation/benchmark-hub)