EmbodiedEvalKit: Unified Evaluation for Embodied AI

• EmbodiedEvalKit is a framework for evaluating embodied AI through formalized tasks, standardized agent interfaces, and unified metric pipelines.
• It unifies diverse systems—like TEA, EmbodiedCity, MFE-ETP, OmniEAR, and Embodied Arena—into a reproducible evaluation workflow.
• The toolkit emphasizes graph-native task representations, modular architectures, and comprehensive metrics to benchmark embodied intelligence.

Searching arXiv for "^^^^^^^^13EmbodiedEvalKit13 and the cited papers to ground the article in the current literature. arxiv_search({"^^^^^^^^13query13 "^^^^^^^^13max_results13 ^^^^13query13EmbodiedEvalKit13^^^^, "^^^^^^^^13sort_by13 arxiv_search({"^^^^^^^^13query13 "^^^^^^^^13max_results13 ^^^^13submittedDate13^^^^, "^^^^^^^^13sort_by13 arxiv_search({"^^^^^^^^13query13 Cognitive Task Generation for In-Situ Evaluation of Embodied Agents^^^^^^^^13query13EmbodiedEvalKit13 "^^^^^^^^13max_results13 ^^^^13submittedDate13^^^^, "^^^^^^^^13sort_by13 arxiv_search({"^^^^^^^^13query13 Arena^^^^^^^^13query13EmbodiedEvalKit13 embodied evaluation platform", "^^^^^^^^13max_results13 ^^^^13submittedDate13^^^^, "^^^^^^^^13sort_by13 ^^^^^^^^13EmbodiedEvalKit13^^^^^^^^ is a recurrent label in recent embodied-AI literature for infrastructures that evaluate agents, models, or embodied experience through standardized task definitions, agent interfaces, and metric pipelines. Current usage suggests that the term does not denote a single canonical package; rather, it names several distinct but overlapping systems, including the TEA in-situ evaluation framework for unseen ^^^^13max_results13^^^^D environments, the EmbodiedCity simulator interface for urban embodied tasks, the MFE-ETP automatic benchmark toolkit for multi-modal foundation models, the OmniEAR reasoning harness for tool use and coordination, the Embodied Arena unified leaderboard platform, and the evaluation framework released with Embodied-R^^^^13query13^^^^.^^^^13submittedDate13^^^^ (&&&^^^^13EmbodiedEvalKit13^^^^&&&, &&&^^^^13EmbodiedEvalKit13^^^^&&&, &&&^^^^13max_results13^^^^&&&, &&&^^^^13sort_by13^^^^&&&, &&&^^^^13submittedDate13^^^^&&&, &&&^^^^13query13^^^^&&&). Across these usages, ^^^^^^^^13EmbodiedEvalKit13^^^^^^^^ denotes a reproducible layer that binds scene or scenario representation, task formalization, model inference, and metric computation into a common evaluation workflow.

^^^^13query13^^^^. Multiple uses of the term in the literature

The name “^^^^^^^^13EmbodiedEvalKit13 appears in several papers with different scopes. In some works it refers to an automatic task-generation system operating inside an unmodified ^^^^13max_results13^^^^D scene; in others it denotes a benchmark SDK, a cross-benchmark orchestration layer, or an evaluation harness for embodied VLMs. This plurality is itself informative: it indicates that embodied-AI evaluation is converging on a toolkit-centric paradigm, but has not yet converged on a single standardized implementation (&&&^^^^13EmbodiedEvalKit13^^^^&&&, &&&^^^^13EmbodiedEvalKit13^^^^&&&, &&&^^^^13max_results13^^^^&&&, &&&^^^^13sort_by13^^^^&&&, &&&^^^^13submittedDate13^^^^&&&, &&&^^^^13query13^^^^&&&).

Paper	Domain	Defining focus
TEA ^^^^^^^^13(He et al., 5 Feb 2026)13	Unseen household ^^^^13max_results13^^^^D environments	Two-stage Interaction/Evolution task generation
EmbodiedCity (&&&^^^^13EmbodiedEvalKit13^^^^&&&)	Real-world city environment	Unreal Engine simulator, AirSim-based access, five benchmark tasks
MFE-ETP (&&&^^^^13max_results13^^^^&&&)	Household embodied task planning	Four-capability hierarchy and automatic evaluation platform
OmniEAR (&&&^^^^13sort_by13^^^^&&&)	Text-based embodied reasoning	Tool usage, continuous properties, implicit collaboration
Embodied Arena (&&&^^^^13submittedDate13^^^^&&&)	Unified cross-benchmark evaluation	^^^^13EmbodiedEvalKit13EmbodiedEvalKit13^^^^ benchmarks, taxonomy, leaderboards, evolving data
Embodied-R^^^^13query13^^^^.^^^^13submittedDate13 (&&&^^^^13query13^^^^&&&)	Embodied VLM benchmarking	Unified grounding/reasoning API and manipulation evaluation

A common structural motif is the separation between data or scene representation, model or agent adapters, and metric computation. Another common motif is formal task definition: tasks are not treated merely as natural-language prompts, but as structured objects with explicit state, goal, and success conditions. This suggests a shift away from narrow benchmark scripts toward reusable evaluation substrates.

^^^^13EmbodiedEvalKit13^^^^. Graph-native task and environment representations

One of the most explicit formalizations appears in the TEA framework, where each task PRESERVED_PLACEHOLDER_^^^^13EmbodiedEvalKit13^^^^ is defined as a pair of labeled graphs, its Initial-State and Final-State:

PRESERVED_PLACEHOLDER_^^^^13query13^^^^

with

PRESERVED_PLACEHOLDER_^^^^13EmbodiedEvalKit13^^^^

Here, PRESERVED_PLACEHOLDER_^^^^13max_results13^^^^ is a set of vertices, PRESERVED_PLACEHOLDER_^^^^13sort_by13^^^^ is a set of edges, and PRESERVED_PLACEHOLDER_^^^^13submittedDate13^^^^ assigns attributes such as object color, ^^^^13max_results13^^^^D bounding box, preconditions, and post-effects. The paper’s abstract “object classification” example instantiates this scheme by adding PRESERVED_PLACEHOLDER_^^^^13query13^^^^ in the Final-State; task instances are then grounded to concrete entities such as “Apple A” ^^^^^^^^13(He et al., 5 Feb 2026)13

TEA organizes generation as a two-stage loop. In the Interaction stage, an agent explores or executes previously generated tasks, collects RGB, depth, instance masks, and poses, and induces new tasks through task generators. In the Evolution stage, the framework applies Task Reuse and Task Recombination to existing graph-structured tasks. Reuse is triggered when PRESERVED_PLACEHOLDER_^^^^13(He et al., 5 Feb 2026)13^^^^, while recombination swaps matching node types across tasks. A diversity filter based on spectral clustering over multi-modal embeddings retains one representative per cluster, with similarity defined as

PRESERVED_PLACEHOLDER_^^^^13max_results13^^^^

The same paper provides concrete examples such as an “object in-view check,” where the graph includes an edge PRESERVED_PLACEHOLDER_^^^^13sort_by13^^^^, and “navigation by label,” where success is formalized as reaching ^^^^13max_results13^^^^D distance PRESERVED_PLACEHOLDER_^^^^13query13EmbodiedEvalKit13^^^^ m from the target within ^^^^13query13EmbodiedEvalKit13^^^^ steps ^^^^^^^^13(He et al., 5 Feb 2026)13

A related but distinct graph formalism appears in OmniEAR. Its text-based simulator EAR-Sim represents the environment as PRESERVED_PLACEHOLDER_^^^^13query13query13^^^^, where PRESERVED_PLACEHOLDER_^^^^13query13EmbodiedEvalKit13^^^^ includes rooms, objects, and agents; PRESERVED_PLACEHOLDER_^^^^13query13max_results13^^^^ stores continuous attributes such as weight, temperature, material composition, dimensions, and dynamic states; and PRESERVED_PLACEHOLDER_^^^^13query13sort_by13^^^^ includes static containment relations and dynamic proximity. Tasks are defined as PRESERVED_PLACEHOLDER_^^^^13query13submittedDate13^^^^, with success determined by whether the final state after applying an action sequence satisfies all predicates in PRESERVED_PLACEHOLDER_^^^^13query13query13^^^^ (&&&^^^^13sort_by13^^^^&&&).

OmniEAR extends this representation to dynamic capability acquisition. Agents begin with a basic action set PRESERVED_PLACEHOLDER_^^^^13query13(He et al., 5 Feb 2026)13^^^^; when an agent grasps a tool object, its available abilities are augmented by that tool’s provides_abilities, and they are removed upon release. Multi-agent collaboration is likewise grounded in physical constraints: when object weight exceeds a single agent’s capacity, the simulator triggers a collaboration protocol that produces a CORP_GRAB → CORP_GOTO → CORP_PLACE sequence without explicit instruction. This suggests a broader trend in ^^^^^^^^13EmbodiedEvalKit13 systems toward graph-native state, explicit affordances, and executable goal predicates rather than prompt-only evaluation (&&&^^^^13sort_by13^^^^&&&).

^^^^13max_results13^^^^. Architectural patterns and system interfaces

At the simulator level, EmbodiedCity presents ^^^^^^^^13EmbodiedEvalKit13^^^^^^^^ as a benchmark platform built in Unreal Engine ^^^^13submittedDate13^^^^.^^^^13max_results13 covering a PRESERVED_PLACEHOLDER_^^^^13query13max_results13^^^^ commercial district in Beijing plus adjacent residential interiors. The environment hosts static geometry as fully textured ^^^^13max_results13^^^^D meshes, with approximately ^^^^13EmbodiedEvalKit13EmbodiedEvalKit13EmbodiedEvalKit13^^^^ buildings, approximately ^^^^13query13EmbodiedEvalKit13EmbodiedEvalKit13^^^^ streets, and more than ^^^^13query13^^^^,^^^^13EmbodiedEvalKit13EmbodiedEvalKit13EmbodiedEvalKit13^^^^ urban assets. Geometry is reconstructed from Baidu and Amap street-view imagery and GIS data; pedestrian and vehicle flows are driven by the Mirage Simulation System using discrete-time kinematics, and access is exposed through a Microsoft AirSim plugin extended with a Python proxy server using HTTP/JSON (&&&^^^^13EmbodiedEvalKit13^^^^&&&).

EmbodiedCity’s agent interface is unusually rich. Per time step it can return RGB images, depth maps, semantic segmentation, IMU, GPS, LiDAR, position, velocity, orientation, and vehicle wheel angle. Action spaces are platform-specific: drones receive position or velocity targets, attitude set-points, camera pan/tilt, and takeoff/land commands, whereas ground vehicles receive steering angle, throttle, brake force, gear shift, and camera controls. The Python SDK supports synchronous and asynchronous calls, batched requests, and live video-stream endpoints for up to eight agents in parallel (&&&^^^^13EmbodiedEvalKit13^^^^&&&).

MFE-ETP presents a more benchmark-centric software stack. Its ^^^^^^^^13EmbodiedEvalKit13^^^^^^^^ is organized into data_loader, prompt_engine, model_adapters, evaluator, and metrics, plus a CLI wrapper. data_loader normalizes raw JSON annotations and image frames into Case(sample_id, images, prompt_text, gt_label), prompt_engine produces templated inputs, model_adapters wrap interfaces such as GPT-^^^^13sort_by13^^^^VAdapter, BLIP^^^^13EmbodiedEvalKit13^^^^Adapter, and MiniGPT^^^^13sort_by13^^^^Adapter, evaluator includes QAEvaluator and HumanEvaluator, and metrics emits per-case and aggregated JSON/CSV outputs. This is a classical end-to-end benchmark harness: dataset preparation, prompt realization, model invocation, scoring, and report export are all explicit modules (&&&^^^^13max_results13^^^^&&&).

Embodied Arena generalizes this pattern into a service-oriented architecture. Its components are a Model Adapter Layer with a uniform predict(input: JSON) → output: JSON interface, a Data Manager using a standard JSON schema, a Scenario Simulator & Orchestrator for interactive tasks in simulators such as Unity and Habitat, a Metric Engine, and a Leaderboard Builder. Internally, benchmarks are wrapped in a common “Task” class, and evaluation is expressed as a loop over benchmarks, cases, predictions, and metric computation. The framework is therefore oriented not only toward single-benchmark experiments but toward unified, leaderboard-scale comparison across heterogeneous embodied tasks (&&&^^^^13submittedDate13^^^^&&&).

The Embodied-R^^^^13query13^^^^.^^^^13submittedDate13^^^^ version of ^^^^^^^^13EmbodiedEvalKit13^^^^^^^^ pushes further toward standardized embodied VLM evaluation. It defines a four-layer stack: a Data Layer in which each benchmark is preprocessed into a single HuggingFace Parquet file with prompt, gt, and meta fields; an Inference Layer with pluggable backends including vLLM, HuggingFace Transformers, and OpenAI/Gemini APIs; a Parsing Layer that normalizes heterogeneous outputs into a ModelOutput dataclass containing text, boxes, points, and trajectories; and an Evaluation Layer implementing each benchmark’s official metric. The framework also includes an envs/ module for optional simulator or real-robot wrapping (&&&^^^^13query13^^^^&&&).

^^^^13sort_by13^^^^. Task families and metric systems

The task taxonomies covered by ^^^^^^^^13EmbodiedEvalKit13 systems span perception, reasoning, navigation, planning, grounding, trajectory prediction, and manipulation. EmbodiedCity defines five benchmark tasks: First-view Scene Understanding, Embodied Question Answering, Embodied Dialogue, Vision-and-Language Navigation, and Embodied Task Planning. MFE-ETP organizes evaluation into four capabilities—Object Understanding, Spatio-Temporal Perception, Task Understanding, and Embodied Reasoning—each decomposed into concrete sub-tasks such as Type Recognition, Property Recognition, Spatial Relations, Temporal Sequencing, Relevant Object Selection, Step Sequence, Goal Completion, and End-to-end Planning. Embodied Arena systematizes the space into three levels—Perception, Reasoning, and Task Execution—with seven core capabilities and ^^^^13EmbodiedEvalKit13submittedDate13^^^^ fine-grained dimensions, while Embodied-R^^^^13query13^^^^.^^^^13submittedDate13^^^^ groups benchmarks into Planning & Correction, Pointing & Affordance, Visual Trace, Spatial Cognition, and Manipulation in Simulation (&&&^^^^13EmbodiedEvalKit13^^^^&&&, &&&^^^^13max_results13^^^^&&&, &&&^^^^13submittedDate13^^^^&&&, &&&^^^^13query13^^^^&&&).

These frameworks also differ in the granularity and semantics of their metrics. EmbodiedCity uses BLEU-^^^^13query13^^^^…^^^^13sort_by13 ROUGE, METEOR, and CIDEr for scene understanding, the same textual metrics plus Sentence-BERT for QA and dialogue, and standard VLN metrics for navigation: Success Rate, Navigation Error, and SPL. Its path-planning discussion explicitly uses A* with PRESERVED_PLACEHOLDER_^^^^13query13sort_by13^^^^, and SPL is given as

PRESERVED_PLACEHOLDER_^^^^13EmbodiedEvalKit13EmbodiedEvalKit13^^^^

This combination reflects a mixed paradigm in which embodied evaluation includes both language generation quality and control performance (&&&^^^^13EmbodiedEvalKit13^^^^&&&).

TEA emphasizes diversity and scene coverage in addition to task success. It defines the Maximum Independent Subset under redundancy threshold PRESERVED_PLACEHOLDER_^^^^13EmbodiedEvalKit13query13^^^^,

PRESERVED_PLACEHOLDER_^^^^13EmbodiedEvalKit13EmbodiedEvalKit13^^^^

and the Maximum Independent Ratio,

PRESERVED_PLACEHOLDER_^^^^13EmbodiedEvalKit13max_results13^^^^

It also reports enclosing volume of object positions, per-axis spread, standard deviations, mean and standard deviation of instance volumes, and task-specific accuracy or mIoU. Navigation-specific metrics include Success Rate, Navigation Gain, Steps Taken, Target Neglect Rate, and Lack of ^^^^13max_results13^^^^D Awareness ^^^^^^^^13(He et al., 5 Feb 2026)13

MFE-ETP adopts the familiar classification metrics Accuracy, Precision, Recall, and PRESERVED_PLACEHOLDER_^^^^13EmbodiedEvalKit13sort_by13^^^^, then adds planning-oriented IoU, Path Efficiency, and Step Efficiency. Embodied Arena uses Exact Matching Accuracy for closed-vocabulary QA, fuzzy matching through CIDEr, BLEU, and ROUGE or LLM-based scorers for open responses, Success Rate and SPL for navigation, and Task Completion Rate for planning; it also defines benchmark-view and capability-view aggregation rules across ^^^^13EmbodiedEvalKit13EmbodiedEvalKit13^^^^ integrated benchmarks. Embodied-R^^^^13query13^^^^.^^^^13submittedDate13^^^^ further expands the metric repertoire with point-distance partial credit for ^^^^13EmbodiedEvalKit13^^^^D pointing tasks, trajectory RMSE, semantic-similarity scoring via a learned LLM reward model with BLEU fallback, and a binary format-check reward enforcing output templates such as <answer>…</answer> (&&&^^^^13max_results13^^^^&&&, &&&^^^^13submittedDate13^^^^&&&, &&&^^^^13query13^^^^&&&).

A plausible implication is that “^^^^^^^^13EmbodiedEvalKit13 increasingly denotes not merely a dataset wrapper, but a metric algebra across heterogeneous output spaces: text answers, boxes, points, trajectories, control traces, and full task executions.

^^^^13submittedDate13^^^^. Empirical findings enabled by these evaluation kits

The TEA framework reports that, across ^^^^13query13EmbodiedEvalKit13^^^^ unseen scenes, it automatically generated ^^^^13max_results13(He et al., 5 Feb 2026)13^^^^,^^^^13max_results13(He et al., 5 Feb 2026)13query13^^^^ tasks in two cycles, and that a ^^^^13query13EmbodiedEvalKit13^^^^% random subset reviewed by human annotators was judged ^^^^13query13EmbodiedEvalKit13EmbodiedEvalKit13^^^^% physically valid, with ^^^^13sort_by13EmbodiedEvalKit13^^^^.^^^^13max_results13 deemed “helpful in daily life” and ^^^^13sort_by13sort_by13^^^^.^^^^13sort_by13 requiring essential cognitive ability. On ^^^^13max_results13sort_by13max_results13^^^^ sampled tasks, human participants scored PRESERVED_PLACEHOLDER_^^^^13EmbodiedEvalKit13submittedDate13^^^^ accuracy on all tasks, whereas state-of-the-art vision-LLMs ranged from ^^^^13EmbodiedEvalKit13^^^^.^^^^13EmbodiedEvalKit13submittedDate13^^^^ on the worst object localization results to ^^^^13EmbodiedEvalKit13^^^^.^^^^13sort_by13sort_by13^^^^ on the best relationship detection results. The reported basic perception gaps were approximately PRESERVED_PLACEHOLDER_^^^^13EmbodiedEvalKit13query13^^^^ in classification and PRESERVED_PLACEHOLDER_^^^^13EmbodiedEvalKit13(He et al., 5 Feb 2026)13^^^^ in localization. In navigation, GPT-^^^^13sort_by13^^^^o achieved Success Rate PRESERVED_PLACEHOLDER_^^^^13EmbodiedEvalKit13max_results13^^^^ and Navigation Gain PRESERVED_PLACEHOLDER_^^^^13EmbodiedEvalKit13sort_by13^^^^, while o^^^^13query13^^^^ reached Success Rate PRESERVED_PLACEHOLDER_^^^^13max_results13EmbodiedEvalKit13^^^^ and Navigation Gain PRESERVED_PLACEHOLDER_^^^^13max_results13query13^^^^, with humans near ^^^^13query13^^^^.^^^^13EmbodiedEvalKit13^^^^ ^^^^^^^^13(He et al., 5 Feb 2026)13

EmbodiedCity evaluates Fuyu-^^^^13max_results13^^^^B, Qwen-VL, Claude ^^^^13max_results13^^^^ Haiku, GPT-^^^^13sort_by13^^^^ Turbo, and GPT-^^^^13sort_by13^^^^o on large urban splits: ^^^^13query13EmbodiedEvalKit13^^^^.^^^^13EmbodiedEvalKit13 scene-understanding cases, ^^^^13submittedDate13EmbodiedEvalKit13^^^^.^^^^13sort_by13 QA questions, ^^^^13query13EmbodiedEvalKit13^^^^.^^^^13query13 dialogues, ^^^^13query13^^^^.^^^^13max_results13 VLN routes, and ^^^^13query13EmbodiedEvalKit13^^^^.^^^^13query13 planning goals. Claude ^^^^13max_results13^^^^ leads scene understanding with CIDEr approximately ^^^^13EmbodiedEvalKit13sort_by13^^^^, GPT-^^^^13sort_by13^^^^ Turbo is reported as close second, GPT-^^^^13sort_by13^^^^ Turbo outperforms the next-best model on QA by approximately ^^^^13query13EmbodiedEvalKit13EmbodiedEvalKit13^^^^% in BLEU/SPL metrics, GPT-^^^^13sort_by13^^^^o reaches approximately ^^^^13(He et al., 5 Feb 2026)13(He et al., 5 Feb 2026)13^^^^% short-route Success Rate and approximately ^^^^13(He et al., 5 Feb 2026)13query13^^^^% SPL in VLN with mean SR approximately ^^^^13max_results13submittedDate13^^^^%, and Claude ^^^^13max_results13^^^^ is best on planning with BLEU-^^^^13sort_by13^^^^ approximately ^^^^13sort_by13^^^^.^^^^13sort_by13^^^^ and CIDEr approximately ^^^^13EmbodiedEvalKit13query13^^^^.^^^^13max_results13^^^^ (&&&^^^^13EmbodiedEvalKit13^^^^&&&).

OmniEAR provides a controlled study of embodied reasoning under constraint-driven planning. Reported single-agent success rates are ^^^^13max_results13submittedDate13^^^^–^^^^13sort_by13query13 for Direct Command, ^^^^13submittedDate13query13^^^^–^^^^13max_results13submittedDate13 for Tool Use, ^^^^13sort_by13EmbodiedEvalKit13^^^^–^^^^13(He et al., 5 Feb 2026)13max_results13 for Attribute Reasoning, and ^^^^13submittedDate13sort_by13^^^^–^^^^13(He et al., 5 Feb 2026)13query13 for Compound Reasoning. Multi-agent success rates are ^^^^13max_results13max_results13^^^^–^^^^13sort_by13EmbodiedEvalKit13 for Explicit Collaboration, ^^^^13query13max_results13^^^^–^^^^13max_results13submittedDate13 for Implicit Collaboration, and ^^^^13EmbodiedEvalKit13max_results13^^^^–^^^^13sort_by13sort_by13 for Compound Collaboration. Fine-tuning Qwen^^^^13EmbodiedEvalKit13^^^^.^^^^13submittedDate13 raises single-agent performance from ^^^^13EmbodiedEvalKit13^^^^.^^^^13query13 to ^^^^13(He et al., 5 Feb 2026)13query13^^^^.^^^^13max_results13 on Direct Command, but improves implicit collaboration only from ^^^^13query13^^^^.^^^^13submittedDate13 to ^^^^13submittedDate13^^^^.^^^^13submittedDate13 An especially notable result is the “Full-Environment-Information Effect”: complete world-graph information improves Tool Use by up to PRESERVED_PLACEHOLDER_^^^^13max_results13EmbodiedEvalKit13^^^^ and Direct Commands by PRESERVED_PLACEHOLDER_^^^^13max_results13max_results13^^^^ to PRESERVED_PLACEHOLDER_^^^^13max_results13sort_by13^^^^, but degrades Implicit Collaboration by ^^^^13submittedDate13^^^^–^^^^13EmbodiedEvalKit13max_results13 (&&&^^^^13sort_by13^^^^&&&).

MFE-ETP reports, at the level of its abstract, that several state-of-the-art MFMs significantly lag behind human-level performance on embodied task planning (&&&^^^^13max_results13^^^^&&&). Embodied Arena extends this diagnosis to a unified leaderboard setting: it reports that object and spatial perception are the weakest links, with average scores approximately ^^^^13EmbodiedEvalKit13max_results13^^^^–^^^^13max_results13max_results13 and that embodied capabilities correlate strongly, with PRESERVED_PLACEHOLDER_^^^^13max_results13submittedDate13^^^^, with downstream navigation and planning performance. It also reports that purely voxel or point-cloud methods underperform by ^^^^13query13submittedDate13^^^^–^^^^13EmbodiedEvalKit13submittedDate13 relative to approaches blending native ^^^^13max_results13^^^^D representations with ^^^^13EmbodiedEvalKit13^^^^D feature alignment (&&&^^^^13submittedDate13^^^^&&&).

The Embodied-R^^^^13query13^^^^.^^^^13submittedDate13^^^^ paper uses its ^^^^^^^^13EmbodiedEvalKit13^^^^^^^^ to evaluate a single ^^^^13max_results13^^^^B EFM across ^^^^13EmbodiedEvalKit13sort_by13^^^^ embodied VLM benchmarks and ^^^^13sort_by13^^^^ manipulation suites, reporting state of the art on ^^^^13query13query13^^^^ out of ^^^^13EmbodiedEvalKit13sort_by13^^^^ embodied VLM benchmarks and strong zero-shot real-robot results in instruction following, affordance grounding, articulated object manipulation, and long-horizon complex tasks (&&&^^^^13query13^^^^&&&). In this context, ^^^^^^^^13EmbodiedEvalKit13^^^^^^^^ functions as the experimental substrate that makes such claims comparable across planning, pointing, spatial cognition, and manipulation.

^^^^13query13^^^^. Limitations, standardization pressures, and adjacent uses

The literature also makes clear that ^^^^^^^^13EmbodiedEvalKit13 infrastructures inherit the limitations of their generators, simulators, and task abstractions. In TEA, reliance on a VLM for semantic tasks introduces hallucination risk and task redundancy; MIR filtering mitigates but does not eliminate redundancy. The framework is evaluated with a simulated Unreal Engine agent rather than a real robot, so physical transfer gaps remain untested, and some high-level tasks such as long-horizon planning are currently out of scope ^^^^^^^^13(He et al., 5 Feb 2026)13

Broader unified platforms surface additional tensions. Embodied Arena argues that models overfit to narrow benchmarks and that isolated tasks fail to reveal comprehensive embodied ability, motivating evolving data-generation methods and monthly updated leaderboards (&&&^^^^13submittedDate13^^^^&&&). OmniEAR identifies architectural bottlenecks in attention-based models, arguing that they struggle to filter task-relevant physical constraints from noisy graphs and to maintain working memory for object states, capability sets, and multi-agent plans. The paper’s own implications for ^^^^^^^^13EmbodiedEvalKit13^^^^^^^^ design include selective constraint filtering modules, persistent state memory or an external graph store, hybrid symbolic-neural architectures for explicit physics reasoning, and modular evaluation kits that expose both abstract planning and continuous constraint reasoning channels (&&&^^^^13sort_by13^^^^&&&).

The term also extends beyond embodied-agent benchmarking into VR embodiment measurement. The Virtual Embodiment Questionnaire defines three factors—ownership, agency, and change in the perceived body schema—validated through confirmatory factor analysis and operationalized as a ^^^^13query13EmbodiedEvalKit13^^^^-statement questionnaire on a seven-point Likert scale. The paper concludes that the VEQ’s sensitivity to first-person immersion and sensorimotor disruption makes it a versatile tool for ^^^^^^^^13EmbodiedEvalKit13 across VR therapy, training, social VR, and entertainment (&&&^^^^13sort_by13query13^^^^&&&). This is not an embodied-agent benchmark in the same sense as TEA or Embodied Arena, but it shows that the “^^^^^^^^13EmbodiedEvalKit13 label can also encompass psychometric evaluation of embodied user experience.

Taken together, the literature presents ^^^^^^^^13EmbodiedEvalKit13^^^^^^^^ not as a settled product category but as an evolving family of evaluation infrastructures. Some instances emphasize in-situ task generation in unseen scenes; others emphasize simulator access, unified benchmark wrapping, capability taxonomies, leaderboard governance, or multi-format metric engines. The common denominator is methodological: embodied intelligence is being evaluated through formalized tasks, explicit interfaces, and increasingly automated data generation rather than through ad hoc benchmark scripts alone.