FLORA: An Open Model Suite for Scaling Robotics Data Engines

Sicong Leng; Yunze Man; Zhiqi Li; Jing Wang; Yuhe Zhang; Ryan Punamiya; Siddharth Patki; Kaiyuan Zheng

cs.ROOct 22, 2026

FLORA: An Open Model Suite for Scaling Robotics Data Engines

Authors: Sicong Leng, Yunze Man, Zhiqi Li, Jing Wang, Yuhe Zhang, Ryan Punamiya, Siddharth Patki, Kaiyuan Zheng,

Abstract

The life-cycle of robotics foundation models crucially depends on data labels. Outcome reward labels are needed for monitoring real-world deployments. Process reward labels are needed during RL post-training. Timestamped captions are needed for filtering and balancing pretraining data. We present Flora, an industrial-quality data engine used for training GR00T models, that simultaneously covers these capabilities. Flora achieves SOTA scores on the captioning benchmark WGO-Bench and reward benchmarks RoboReward and Robometer. On internal benchmarks it is competitive with GPT-6 Astra while beating other frontier models. The model is substantially cheaper than API-based alternatives, roughly 125 times cheaper than GPT-6 Astra. Florais battle tested at scale, labeling 200,000 hours of data per day across 512 GB300s. By open-sourcing our data engine, we enable researchers to scale custom robotics datasets and explore emerging applications such as physical auto-research and human-to-robot transfer.

Figures & tables

Figure 1: Capabilities. Flora produces timestamped captions by first segmenting videos and then captioning individual segments. Flora can also produce outcome and process rewards, and rewards based on rubrics specified at test-time.

Figure 2: Annotation quality versus cost. Flora achieves competitive quality at substantially lower cost compared to GPT-6 Astra. Evaluation and cost assumptions appear in Appendix D .

Table 1: Benchmark scores. Flora is competitive with GPT-6 Astra while beating other baselines. Values are scaled by 100. Full comparisons appear in Appendix A .

Figure 3: One data engine, complementary capabilities. Timestamped caption and reward labeling results across six tasks. All values are scaled by 100. Full comparisons appear in Appendix A .

Figure 4: Complete pipelines on WGO-Bench. Equal-subset mean scores (%, $\uparrow$ ), with generation backbones shown. Full results and protocols appear in Appendix B.6 .

Figure 5: From intended actions to execution quality. Selected examples show temporal coverage (top), human action captions (middle), and failure recognition and partial progress on DenseRewardBench and RoboRewardBench (bottom).

Figure 6: From video to reusable supervision. Overlapping windows are stitched into event intervals for contextual captioning. One reward model provides dense, outcome, or rubric labels. Timelines are schematic.

Appendix figures & tables18 assets

Supplementary material from the paper’s appendix.

Appendix

	Segmentation [email protected]				Captioning Gold-window accuracy				Joint annotation Semantic [email protected]
Model / system	HomER	DROID	Galaxea	Avg	HomER	DROID	Galaxea	Avg	HomER	DROID	Galaxea	Avg
Proprietary models
GPT-6 Astra	25.2	65.5	47.2	46.0	64.0	82.0	86.2	77.4	19.3	54.5	43.1	39.0
GPT-5.6 Sol	23.5	44.0	33.5	33.7	48.7	62.7	58.5	56.6	16.9	33.6	23.4	24.6
GPT-5.6 Luna	20.6	32.3	37.8	30.2	46.8	55.3	57.7	53.3	13.3	22.3	24.0	19.9
Gemini 3.8 Flash	23.9	30.8	24.5	26.4	63.0	70.0	79.7	70.9	20.6	23.7	20.9	21.7

Appendix

Table 2: WGO-Bench results. Subset scores and equal-weight averages (%,

\uparrow

). Off-the-shelf models use the common pipeline , and Flora uses its default pipeline (Section 4.1 ).

Model / system	Segmentation [email protected]				Oracle captioning Point quality				Joint annotation Point [email protected]
	Human	Robot			Human	Robot			Human	Robot
		Grip.	Dex.	Avg		Grip.	Dex.	Avg		Grip.	Dex.	Avg
Proprietary models
GPT-6 Astra	44.2	70.7	73.1	71.9	60.3	79.6	80.8	80.2	29.7	57.7	58.1	57.9
GPT-5.6 Sol	32.6	32.4	59.2	45.8	36.4	68.8	70.3	69.6	14.6	22.8	44.3	33.6
GPT-5.6 Luna	21.4	23.8	51.9	37.9	28.5	54.2	52.1	53.2	8.1	13.7	31.5	22.6

Appendix

Table 3: DenseCaptionBench results. Scores on 100 human, 100 parallel-gripper (Grip.), and 100 dexterous-hand (Dex.) episodes under the shared point-quality protocol (%,

\uparrow

). Off-the-shelf models use the common pipeline , and Flora uses its production pipeline (Section 4.1 ).

	RoboRewardBench	RobometerBench
Model	MAE $\downarrow$	OOD Kendall $\tau_{a}$ $\uparrow$	ID VOC $\uparrow$	OOD VOC $\uparrow$
Proprietary models
GPT-6 Astra	55.4	82.3	98.9	99.7
GPT-5.6 Sol	78.0	64.2	97.1	99.3
GPT-5.6 Luna	77.9	59.3	95.7	98.7
Gemini 3.8 Flash	83.3	63.3	96.9	98.6

Appendix

Table 4: Public reward benchmarks. Flora-Rewarder has the lowest reported MAE and exceeds the specialized reward baselines on their reported metrics. Published specialized-model scores retain their original protocols. All metrics are multiplied by 100.

	DenseRewardBench				RubricBench
Model	Macro F1 $\uparrow$	Balanced acc. $\uparrow$	Bad-action F1 $\uparrow$	Failure recall $\uparrow$	Pooled L1 $\downarrow$
Proprietary models
GPT-6 Astra	40.7	44.1	44.8	0.0	3.6
GPT-5.6 Sol	36.3	37.8	26.4	1.0	8.4
GPT-5.6 Luna	34.7	34.9	17.6	2.3	22.6
Gemini 3.8 Flash	38.8	40.2	31.2	11.7	32.6

Appendix

Table 5: Internal reward benchmarks. Flora-Rewarder achieves the best value on all four DenseRewardBench metrics and grades rubrics more accurately than nine of the ten general-purpose models. All metrics are multiplied by 100.

Embodiment	Visual evidence	Episodes	References
Human	Egocentric camera	100	3,220
Parallel grippers	Main and two wrist cameras	100	1,266
Dexterous hands	Main and two wrist cameras	100	391
Total		300	4,877

Appendix

Table 6: DenseCaptionBench composition. Reference intervals may include idle gaps and concurrent intended actions.

		Output limit
Model	Reasoning and sampling	DenseCaptionBench	WGO
GPT-6 Astra	xhigh , default sampling	32k	32k
GPT-5.6 Sol / Luna	xhigh , default sampling	32k	32k
Gemini 3.8 Flash	Default reasoning, temperature 0	32k	32k
Gemini 3.1 Pro	Default reasoning, temperature 0	16k	32k
Claude Opus 5 / Sonnet 5	Provider defaults	32k	32k

Appendix

Table 7: Baseline decoding configurations. Output limits are in thousands of tokens (k = 1,024). Default denotes an omitted request field, not a verified reasoning level.

	Segmentation [email protected]				Captioning Gold-window accuracy				Joint annotation Semantic [email protected]
Pipeline / backbone	HomER	DROID	Galaxea	Avg	HomER	DROID	Galaxea	Avg	HomER	DROID	Galaxea	Avg
Shared backbone: Gemini 3.5 Flash [ 36 ]
Common pipeline	13.2	28.6	34.0	25.3	55.7	56.7	59.3	57.2	10.7	19.7	26.4	18.9
Refiner [ 2026b ]	22.0	31.4	46.9	33.4	57.7	65.3	69.9	64.3	15.8	26.5	39.4	27.2
LeRobot annotate [ 2026 ]	14.2	28.3	31.0	24.5	47.9	34.7	41.5	41.4	8.0	20.3	22.9	17.1
Model-specific backbones

Appendix

Table 8: WGO-Bench pipeline comparison. Subset scores and equal-weight averages (%,

\uparrow

). Subsequent off-the-shelf model comparisons use the common pipeline as a simple reference protocol that standardizes input construction and prompts while limiting API-call and visual-token overhead.

Pipeline	Backbone	Subset	Segmentation [email protected]	Captioning Accuracy	Joint [email protected]
Common pipeline	Gemini 3.8 Flash	HomER	23.9	63.0	20.6
		DROID	30.8	70.0	23.7
		Galaxea	24.5	79.7	20.9
		Avg	26.4	70.9	21.7
Refiner	Gemini 3.8 Flash	HomER	26.3	60.0	20.7
		DROID	48.2	72.7	41.0

Appendix

Table 9: Pipeline comparison with Gemini 3.8 Flash. Scores are percentages (

\uparrow

). Avg equally weights the three WGO subsets. Green numbers identify Flora variants. Shaded cells mark the best averages, including ties. The Pegasus and trained Flora results match Figure 4 .

Models	RPM	Input/admission	Output
GPT-6 Astra / GPT-5.6 Sol	15,000	40M	Included in admission proxy
Claude Opus 5 / Sonnet 5	10,000	10M	2M, including reasoning
Gemini 3.8 Flash	20,000	20M	No separate allowance modeled
Gemini 3.1 Pro	2,000	8M	No separate allowance modeled

Appendix

Table 10: Quota scenarios for throughput estimates. RPM denotes requests per minute. M denotes million tokens per minute.

Figure 7: Annotation quality versus conditional throughput. The 50-episode quality scores match Figure 2 . API rates follow the quota scenarios in Table 10 . The Flora rate assumes 12 independent eight-GPU replicas. Vertical bars are episode-bootstrap 95% intervals. Throughput is modeled, not measured, and the points are not budget matched.

Figure 8: Actions and rewards over time. Three consecutive actions (top) and a localized grasp failure (bottom), from two separate episodes.

Figure 9: Example episode outputs across embodiments. Timestamped intended-action captions appear above dense execution-quality and outcome labels. Robot mosaics retain synchronized scene and wrist views. Selected captions and rationales illustrate the structured records generated by the engine.

Figure 10: Qualitative comparison with GPT-6 Astra. Flora captures better dense actions (a) and local manipulation details (c,e), but can over-segment coarse events (b) and confuse action phases or objects (d,f). Parentheses denote segment counts. Temporal and caption scores are F1@0.5 and point quality on reference intervals, respectively. Robot insets show right-wrist views. — Figure 10: Qualitative comparison with GPT-6 Astra. Flora captures better dense actions (a) and local manipulation details (c,e), but can over-segment coarse events (b) and confuse action phases or objects (d,f). Parentheses denote segment counts. Temporal and caption scores are [email protected] and point quality on reference intervals, respectively. Robot insets show right-wrist views.

Figure 11: Flora-Rewarder flags failures and credits progress that GPT-6 Astra misjudges. Selected examples. (a–d) DenseRewardBench: Astra calls a falling wing and a missed grasp suboptimal, although its own rationales describe both. (e,f) RoboRewardBench: Astra scores a partial and a full success too low. (a,b) 50 s excerpts with per-second labels. (c,d) the dashed 16 s windows. Right-hand numbers give agreement with the reference (%).

Figure 12: Windowed localization and contextual captioning. Overlapping windows emit fragments from their central regions. Timestamp projection and stitching reconstruct episode intervals, which are captioned independently with before, target, and after visual evidence.

Figure 13: Synchronized views and temporal context. Each mosaic aligns cameras at a source time. Mosaics retain their temporal order within the before, target, and after blocks.

Figure 14: Three interfaces, one reward model. Outcome prompts score episode completion, dense prompts label temporal groups with rationales, and rubric prompts assign discrete credit to user-defined task steps.

Step	Score	Criterion
1. Firmly grasp apple	0.0	Did not grasp the apple
	0.2	Touched but did not grab the apple
	0.4	Firmly grasped the apple
2. Pick up knife	0.0	Did not pick up the knife
	0.15	Grasped the knife but dropped it
	0.3	Picked up the knife with a secure grip

Appendix

Table 11: Sample scoring rubric for an apple-cutting task.

Robot Foundation Models

Robot Foundation Models are pretrained systems that combine perception, reasoning, and action to support diverse tasks and robot embodiments.

See all 135 papers

Robometer: Scaling General-Purpose Robotic Reward Models via Trajectory Comparisons
Mar 2, 2026Anthony Liang, Yigit Korkmaz, Jiahui Zhang +14Reward Modeling
Univ. of Southern California · UT Dallas · MIT +4
Rufus-Air: An Open LLM Post-Training Recipe
Sep 24, 2026Chia-Yuan Chang, Renyuan Cheng, Rui Feng +19LLM Fine-Tuning RL for Language Model Reasoning
Amazon
Scaling Vision-Language Reward Learning for Robot Manipulation in Parallel Simulation
Sep 18, 2026Lobna Joualy, Eric Demeester, Nikolaos TsiogkasRobotic RL Preference-Based RL
Department of Computer Science and Department of Mechanical Engineering, KU Leuven, 3590 Diepenbeek, Belgium.

Mar 2, 2026J/K move · Enter open · S save

Robometer: Scaling General-Purpose Robotic Reward Models via Trajectory Comparisons

Anthony Liang, Yigit Korkmaz, Jiahui Zhang, Minyoung Hwang+13

Univ. of Southern California · UT Dallas · MIT+4

Reward Modeling

Open paper AskPDF arXiv

General-purpose robot reward models are typically trained to predict absolute task progress from expert demonstrations, providing only local, frame-level supervision. While effective for expert demonstrations, this paradigm scales poorly to large-scale robotics datasets where failed and suboptimal trajectories are abundant and assigning dense progress labels is ambiguous. We introduce Robometer, a scalable reward modeling framework that combines intra-trajectory progress supervision with inter-trajectory preference supervision. Robometer is trained with a dual objective: a frame-level progress loss that anchors reward magnitude on expert data, and a trajectory-comparison preference loss that imposes global ordering constraints across trajectories of the same task, enabling effective learning from both real and augmented failed trajectories. To support this formulation at scale, we curate RBM-1M, a reward-learning dataset comprising over one million trajectories spanning diverse robot embodiments and tasks, including substantial suboptimal and failure data. Across benchmarks and real-world evaluations, Robometer learns more generalizable reward functions than prior methods and improves robot learning performance across a diverse set of downstream applications. Code, model weights, and videos at https://robometer.github.io/.

Abstract

Robot Foundation Models

Explore similar work

Robometer: Scaling General-Purpose Robotic Reward Models via Trajectory Comparisons

Rufus-Air: An Open LLM Post-Training Recipe

Scaling Vision-Language Reward Learning for Robot Manipulation in Parallel Simulation