RoboFollow: Unveiling the Instruction Following Mirage in Embodied Agents

Hugging Face Daily Papers Papers
embodied-agents instruction-following benchmark ai-research robotics language-guidance evaluation

Summary

RoboFollow introduces a diagnostic benchmark to expose the illusion of instruction-following in embodied agents by analyzing high scene entropy and perturbations, revealing gaps in current models despite strong initial performance.

Modern embodied agents achieve impressive success rates, yet their actual instruction-following ability is far weaker than these numbers suggest. We trace this illusion to a structural property we term low scene entropy: when a visual scene admits only one valid task, language becomes redundant and a policy can score highly while barely using it. We introduce RoboFollow, a diagnostic benchmark with three principles: (1) High Scene Entropy: each training scene supports multiple kinematically distinct task branches, making vision alone insufficient and forcing reliance on language. (2) Hierarchical Diagnostic Protocol: a four-level protocol (L0--L3) progressively perturbs visual layout and semantics, probing whether equivalent instructions yield consistent behavior and distinct ones yield discriminable behavior across spatial relations, attributes, trajectory constraints, and logic. (3) Confound-Controlled Diagnosis: we simplify interaction objects, restrict actions to the trained repertoire and report stage-wise Intent and Execution scores, isolating comprehension from motor execution. Evaluation of nine VLA and WAM policies shows that strong L0 performance, where attained, does not reliably transfer to L1--L3 under our fine-tuning setup. Representative mitigations, including stronger VLM backbones, QA co-training, LangForce, and Classifier-Free Guidance, all fail to close this gap. RoboFollow exposes genuine instruction following as a critical, overlooked bottleneck. Code and dataset are available at https://github.com/AutoLab-SAI-SJTU/RoboFollow and https://huggingface.co/datasets/AutoLab-SJTU/robofollow-data.
Original Article
View Cached Full Text

Cached at: 09/23/26, 11:32 AM

Paper page - RoboFollow: Unveiling the Instruction Following Mirage in Embodied Agents

Source: https://huggingface.co/papers/2609.25636

Abstract

Modernembodiedagentsachieveimpressivesuccessrates,yettheiractualinstruction-followingabilityisfarweakerthanthesenumberssuggest.Wetracethisillusiontoastructuralpropertywetermlowsceneentropy:whenavisualsceneadmitsonlyonevalidtask,languagebecomesredundantandapolicycanscorehighlywhilebarelyusingit.WeintroduceRoboFollow,adiagnosticbenchmarkwiththreeprinciples:(1)HighSceneEntropy:eachtrainingscenesupportsmultiplekinematicallydistincttaskbranches,makingvisionaloneinsufficientandforcingrelianceonlanguage.(2)HierarchicalDiagnosticProtocol:afour-levelprotocol(L0--L3)progressivelyperturbsvisuallayoutandsemantics,probingwhetherequivalentinstructionsyieldconsistentbehavioranddistinctonesyielddiscriminablebehavioracrossspatialrelations,attributes,trajectoryconstraints,andlogic.(3)Confound-ControlledDiagnosis:wesimplifyinteractionobjects,restrictactionstothetrainedrepertoireandreportstage-wiseIntentandExecutionscores,isolatingcomprehensionfrommotorexecution.EvaluationofnineVLAandWAMpoliciesshowsthatstrongL0performance,whereattained,doesnotreliablytransfertoL1--L3underourfine-tuningsetup.Representativemitigations,includingstrongerVLMbackbones,QAco-training,LangForce,andClassifier-FreeGuidance,allfailtoclosethisgap.RoboFollowexposesgenuineinstructionfollowingasacritical,overlookedbottleneck.Codeanddatasetareavailableathttps://github.com/AutoLab-SAI-SJTU/RoboFollowandhttps://huggingface.co/datasets/AutoLab-SJTU/robofollow-data.

View arXiv pageView PDFProject pageGitHub2Add to collection

Get this paper in your agent:

hf papers read 2609\.25636

Don’t have the latest CLI?curl \-LsSf https://hf\.co/cli/install\.sh \| bash

Models citing this paper0

No model linking this paper

Cite arxiv.org/abs/2609.25636 in a model README.md to link it from this page.

Datasets citing this paper1

#### AutoLab-SJTU/robofollow-data Updatedabout 2 hours ago • 1.39k • 1

Spaces citing this paper0

No Space linking this paper

Cite arxiv.org/abs/2609.25636 in a Space README.md to link it from this page.

Collections including this paper0

No Collection including this paper

Add this paper to acollectionto link it from this page.

Similar Articles

ROGUE: Misaligned Agent Behavior Arising from Ordinary Computer Use

arXiv cs.LG

This paper introduces ROGUE, a benchmark to evaluate corrigibility failures in AI agents, finding that frontier models often bypass user interruptions or restrictions even in benign settings, and that better performance correlates with greater misalignment.