---
title: 'MP1 Algorithm: One-Step Generative Policy Learning'
url: https://www.emergentmind.com/topics/mp1-algorithm
type: topic
---

# MP1 Algorithm: One-Step Generative Policy Learning

MP1 is a generative policy learning algorithm for robotic manipulation that leverages the MeanFlow paradigm to deliver one-step (1-NFE) trajectory generation for high-dimensional, context-rich policy inference using 3D point cloud observations and robot state histories. It addresses the trade-off between the slow, autoregressive sampling of diffusion-based policies and the consistency constraints necessary for classical flow-based policies by introducing a new formulation—interval-averaged velocity learning—that is solved efficiently using a U-Net–based model and a dispersive regularizer for few-shot generalization [2507.10543].

## 1. Theoretical Foundations: MeanFlow and the MeanFlow Identity

Traditional Flow Matching learns the instantaneous velocity field $v(z_t, t)$ under the continuity equation $\frac{dz_t}{dt} = v(z_t, t)$, requiring the solution of an ODE during inference. MP1 instead models the *interval-averaged* velocity:
$$
u(z_t, r, t) \triangleq \frac{1}{t - r} \int_r^t v(z_\tau, \tau) \, d\tau
$$
Rather than estimating $v$ pointwise and integrating, MP1 exploits the *MeanFlow Identity*:
$$
u(z_t, r, t) = v(z_t, t) - (t - r)\frac{d}{dt}u(z_t, r, t)
$$
where the total derivative is $\frac{d}{dt}u = v(z_t, t)\,\partial_z u + \partial_t u$. This relates the interval average $u$ to the local velocity, enabling direct, closed-form trajectory prediction without iterated integration or structural consistency losses. The approach eliminates numerical ODE-solver errors at inference.

## 2. Policy Architecture and One-Step Inference

The MP1 network receives:
- A sequence of raw point-clouds $P \in \mathbb{R}^{n_o \times n_p \times 3}$
- A sequence of proprioceptive robot states $S \in \mathbb{R}^{n_o \times s_d}$

Separate encoders extract visual ($f_v$) and state ($f_s$) features, which are concatenated as the conditional code $c = (f_v, f_s)$. The downstream network is a U-Net, which, for given noisy trajectory $A_t$, interval endpoints $r, t$, and condition $c$, predicts $u_\theta(A_t, r, t \mid c)$.

At inference, with $(r, t) = (0, 1)$ and a single Gaussian sample $A_1 \sim \mathcal N(0, I)$,
$$
A_0 = A_1 - u_\theta^{\mathrm{cfg}}(A_1, 0, 1 \mid c)
$$
This constitutes "true" 1-NFE (one network function evaluation): a single forward pass suffices, and no numerical integration is required.

## 3. Training Objective: Classifier-Free Guidance and Dispersive Loss

MP1 uses two primary objective terms:

(a) **CFG regression loss**:
$$
\mathcal{L}_{\mathrm{cfg}}(\theta) = \mathbb{E}_{t, r, A_0, \epsilon} \left\| u_\theta^{\mathrm{cfg}}(A_t, r, t \mid c) - \mathrm{stopgrad}\left(u_{\mathrm{tgt}}\right) \right\|_2^2
$$
with $A_t = (1-t)A_0 + t\epsilon$, and
$$
u_{\mathrm{tgt}} = v_t - (t - r)[v_t\,\partial_z u_\theta + \partial_t u_\theta]
$$

**Classifier-Free Guidance** (CFG) is incorporated by randomly dropping the conditioning $c$ with some probability:
$$
\tilde v_t = \omega\,v_t(A_t \mid A_0, c) + (1-\omega)\,u_\theta^{\text{uncond}}(A_t)
$$
which is used in $u_{\mathrm{tgt}}$ for regression.

(b) **Dispersive Loss**:
$$
\mathcal{L}_{\mathrm{Disp}}(\theta) = \log\,\mathbb{E}_{i \neq j \in \mathcal{B}} \left[ \exp\left( -\frac{\|z_i - z_j\|^2}{\tau} \right) \right],\quad \tau=1
$$
where $z_i, z_j$ are down-block latent representations for different batch samples. This repels embedding vectors in latent space, improving generalization in few-shot learning and discouraging latent collapse.

The total objective is:
$$
\mathcal{L}_{\text{total}}(\theta) = \mathcal{L}_{\mathrm{cfg}}(\theta) + \lambda \mathcal{L}_{\mathrm{Disp}}(\theta),\quad \lambda = 0.5
$$

## 4. Empirical Performance and Ablation Results

On Adroit and Meta-World benchmarks, MP1 outperforms both DP3 (diffusion-based, 10 NFE) and FlowPolicy (flow-based, 1 NFE) in both average success rate and latency:

| Method      | NFE | Avg. Success (\%) | Avg. Inference Time (ms) |
|-------------|-----|-------------------|--------------------------|
| DP3         | 10  | 68.7              | 132.2                    |
| FlowPolicy  | 1   | 71.6              | 12.6                     |
| **MP1**     | 1   | **78.9**          | **6.8**                  |

Ablation studies show:
- Removing Dispersive Loss reduces average success by ~5%.
- Performance is maximized at intermediate interval ratios $r < t$ (not $r = t$, i.e., not classical flow matching), confirming the benefit of interval-averaged flows.
- MP1 maintains superior performance even in few-shot imitation regimes (2–5 demonstrations), with near-state-of-the-art results by 10 demonstrations, attributable to improved discriminative capacity in the latent space.

In real-world robotic manipulation tasks on the ARX R5 dual-arm, MP1 achieves the highest task success rates and fastest average completion times in comparison to both baselines.

## 5. Comparison with Alternative Generative Policy Methods

MP1 achieves:
- 19x reduction in inference latency versus diffusion-based DP3, with improved success rates ($+10.2\%$).
- Nearly 2x faster inference and $+7.3\%$ higher success than the explicit flow-based FlowPolicy.
- True one-step policy generation, with no consistency loss or numerical ODE-solver artifacts.

This is enabled by its local learning strategy (via the MeanFlow Identity), one-step trajectory computation (1-NFE), and lightweight, batch-wide dispersive regularization.

## 6. Implementation and Evaluation Protocols

The reported implementation uses:
- Batch size 128, AdamW $(\text{lr} = 10^{-4})$
- Farthest-point sampling for 512 or 1024 points
- Downsampled images to $84 \times 84$
- Training for 3000 epochs (Adroit) or 1000 (Meta-World), with evaluation every 200 epochs, and top 5 checkpoint selection per seed.

Evaluation comprises 10 expert demos per task in benchmarks, 20 human demos per task in real-world, and assessment on five real tasks. MP1 is deployed with a U-Net backbone and 3D-conditioned inputs on NVIDIA RTX4090 hardware.

## 7. Context, Limitations, and Prospective Directions

By integrating the MeanFlow identity into the policy learning process, MP1 bypasses the need for multi-step iterative sampling and structural consistency. This permits both high-frequency control and robust generalization when only a few demonstrations are available. The main empirical findings demonstrate not only improved task performance but a substantial drop in inference time, which is critical for real-world closed-loop robotic control.

A plausible implication is that interval-averaged flow architectures with dispersive regularization will become the design of choice for settings where latency, sample efficiency, and robust few-shot generalization are all priorities [2507.10543]. The need for only a single function evaluation at deployment may facilitate embedded and resource-constrained robotic systems. However, the long-term impact and generalization beyond the tested benchmarks will depend on continued evaluation across more heterogeneous tasks and sensor modalities.

Source: https://www.emergentmind.com/topics/mp1-algorithm