RynnValue-4B-Quantile

RynnValue-4B-Quantile uses RynnBrain-4B and quantile bins to score a video trajectory against a task instruction. The heads estimate time remaining and signed time displacement. DAMO reports a mean Kendall τₐ of 0.679 on RBM-EVAL-OOD policy ranking with a causal mask.

Papers & reports (1)

RynnValue: Scaling Robotic Value Foundation Models with Temporal Distance

Paper available

arXiv: 2608.09853
Authors & affiliations (15)

Author order: Original order

  1. Dongchi Huangequal_contribution

    Not provided

  2. Hongyin Zhangequal_contribution

    Not provided

  3. Bohan Houequal_contribution

    Not provided

  4. Siteng Huangcorresponding_author

    Not provided

  5. Zhian Su

    Not provided

  6. Hang Guo

    Not provided

  7. Tong Lu

    Not provided

  8. Zhaofeng Xu

    Not provided

  9. Jiahao Tang

    Not provided

First published
2026-08-10
Latest paper revision
v2 · 2026-09-29
Venue
Not provided

Covered model versions

Paper ↗PDF ↗

How to access and use

Public code resources
Public access
Model files
Public access

Training code for this version is unconfirmed. Public code resources do not establish access to the exact training implementation.

Training / fine-tuning code

Model files

Model files
RynnValue-4B-Quantile (native)AccessWeights licenseCommercial use
ConditionalTerms

Verified on · 2026-11-19

Conditions

Apache-2.0 applies. When redistributing, include the license and retain applicable notices.

Release information

SourceVerified on · 2026-11-19

Checks before deployment

Available weights or code do not establish plug-and-play robot compatibility. Check the required configuration, data and interfaces in the source documentation.

Deployment and usage resources

Runtime and frameworks
PyTorch / Transformers

Architecture and inputs / outputs

Model type
Reward & evaluation
Architecture
Transformer
Backbone
RynnBrain-4B (Qwen3-VL)
Weight parameters
5.145 B
Input
Language / Video

Scope and training

Implementation source

Official

Functional purpose

reward_progress

Output

Rewards / progress

Task scope

Multiple tasks

Submit a source or correction