Living Room Tidying: The Home Robot Benchmark

That sounds ordinary. For robots, it is brutal.

Figure's Helix 02 living-room-tidy demo is interesting because it names the right problem. Figure says a living room changes constantly: objects are scattered unpredictably, furniture creates narrow paths, soft objects deform, many actions require both hands, and nearly every behavior mixes walking with manipulation. In the demo, the robot wipes a surface, handles a towel, scoops blocks into a bin, tucks a container under one arm, tosses a pillow back onto the couch, reorients a remote, presses a button, and side-steps through a tight gap.

The right takeaway is not "home chores are solved." They are not. The useful takeaway is that living-room tidying gives buyers a better test than most humanoid hype videos. It asks whether a robot can perceive clutter, decide where things belong, move through a real room, handle soft and rigid objects, recover from mistakes, and keep the task going without a person babysitting every step.

Why tidying is harder than cleaning

Robot vacuums made home robotics familiar because floor cleaning is bounded. The robot mostly stays on the floor. The goal is measurable. The world can be mapped. Even when the robot gets stuck, the problem is usually navigation, suction, mopping, docking, or obstacle avoidance.

Tidying is different. A tidying robot has to answer questions a vacuum never sees:

• Is this object trash, a toy, a book, clothing, or something valuable?
• Does it belong on the table, in a basket, on the couch, in a drawer, or exactly where it is?
• Can I pick it up without crushing it, spilling it, tangling it, or knocking over something nearby?
• Do I need one hand, two hands, a tool, or a temporary place to stash it?
• What should I do if the object slips, folds, rolls, or is partly hidden?

That is why a living-room benchmark matters. It combines six hard problems at once: scene understanding, whole-room navigation, dexterous grasping, soft-object handling, task memory, and error recovery. A robot that only performs one of those well can look impressive in a short clip and still fail the chore.

Figure's own technical framing points in the same direction. In its Helix 02 post, the company described a single neural system controlling the full body from pixels, with vision, touch, and proprioception feeding whole-body control. It also emphasized long-horizon loco-manipulation: walking while holding objects, adjusting balance while reaching, and sequencing many actions over minutes rather than seconds.

That is exactly what a buyer should care about. Not whether a humanoid can stand near a sofa. Whether it can make progress in a room that refuses to stay neat.

The ui44 living-room-tidy checklist

A practical living-room benchmark should not be a single perfect demo. It should be a checklist that separates useful capability from theater.

The last point is underrated. A useful home robot does not need to be proud. If it is unsure whether a wallet belongs in a drawer or on the coffee table, the correct behavior may be to ask, remember the answer, and improve next time. The best living-room benchmark rewards that honesty.

Which robots look closest today?

The ui44 robot database shows a wide gap between "can manipulate objects" and "can tidy a living room as a product." The strongest candidates fall into four groups.

1. Full humanoids with serious manipulation demos

Figure 03 is the clearest example of the direction. In the database it is listed as an active humanoid with a 168 cm body, 60 kg weight, roughly five-hour runtime, 4.3 km/h speed, tactile arrays, force sensors, depth cameras, and Figure's in-house Helix VLA system. The catch is just as important: no price is announced and it is not available as a consumer product.

Figure 02, the robot associated with the public Helix lineage, shows why the hardware class matters. ui44 lists it at 168 cm, 70 kg, with a 20 kg payload and 16-DOF hands, but also marks it discontinued and industrial-only. That makes it valuable evidence for the benchmark, not something a household can buy.

AGIBOT X2 is a different signal: it is actually listed as available, with an official $24,240 price, a 131 cm body, 35-39 kg weight depending on version, roughly two hours of walking at 0.5 m/s, 1.8 m/s top speed, and up to 3 kg payload in specific postures. It has object manipulation with an OmniHand accessory, but the database does not treat it as a proven living-room chore robot. The distinction matters. Availability is not the same as household autonomy.

1X NEO is more home-focused. ui44 lists it as a $20,000 pre-order robot, 167 cm tall, 30 kg, with about four hours of battery life, tactile skin, depth sensing, and capabilities such as household chores, tidying up, safe human interaction, adaptive learning, and gentle manipulation. 1X also describes an "Expert Mode" for chores NEO does not yet know, where a human expert can guide it. That is a realistic bridge to usefulness, but it is also a privacy and reliability boundary buyers should understand.

2. Home-first robots trained on chores

Sunday Memo is one of the most directly relevant entries because its official positioning is about household chores rather than general humanoid spectacle. ui44 tracks Memo as a development-stage home assistant with no announced retail price, a late-2026 limited beta, and capabilities including autonomous table clearing, dishwasher loading, laundry folding, coffee preparation, voice-directed scheduling, app scheduling, and mobile household navigation.

Sunday's official site says Memo is trained with a Skill Capture Glove and that the company has shipped more than 2,000 gloves to people collecting household task data. That is exactly the kind of data pipeline living-room tidying needs. A robot cannot be trained only on clean lab surfaces and then be expected to understand a family's messy den.

Syncere Lume takes the opposite hardware approach. It is not a walking humanoid; it is a sculptural floor lamp with a hidden robotic arm. ui44 lists it as a $1,499 pre-order home assistant, with soft-material chores, bed making, pillow resetting, laundry folding, and one-arm pick-and-place tidying. That limited form factor may actually be sensible for some rooms. If the job is resetting pillows and handling fabric within a known workspace, a fixed or semi-fixed robot can be safer and cheaper than a biped.

3. Mobile manipulators that admit the problem

Hello Robot Stretch 3 is not marketed as a mass-market housekeeper, but it is one of the most honest platforms for the problem. ui44 lists it at $24,950, 24.5 kg, 33 × 34 × 141 cm, with a two-to-five-hour runtime, a 2 kg payload, a telescoping arm, ROS 2, a Python SDK, RGB-D cameras, LiDAR, and web/gamepad/dexterous teleoperation.

That spec sheet explains why research mobile manipulators remain important. Stretch does not need to look human to reach from the floor to a cabinet, navigate tight spaces, collect data, and test autonomy. For living-room tidying, a compact wheeled base plus one good arm may beat a humanoid that walks beautifully but cannot grasp reliably.

Toyota's Human Support Robot is another reminder that this problem is older than the current humanoid boom. Toyota framed HSR around independent living for elderly and disabled users: picking objects up from the floor, retrieving items from shelves, and allowing remote operation by family or caregivers. The database lists a 37 kg body, 100.5-135 cm adjustable height, 0.8 km/h speed, and a 1.2 kg object limit. It is a research/developer platform, not a retail living-room tidier, but its tasks are still the right tasks.

4. Limited consumer robots that tidy one slice

Roborock Saros Z70 is useful because it shows the first consumer-friendly slice of tidying: removing small obstacles before cleaning. ui44 lists it as an available $1,299.99 robot vacuum with a foldable five-axis OmniGrip arm, object pickup for socks, shoes, and small items, AI object recognition, 22,000 Pa suction, vacuuming, mopping, and a multifunction dock.

That is not a general living-room robot. It will not decide where the TV remote belongs or fold a blanket. But it is commercially important because it turns "tidying" into a narrow, shippable feature: move the sock so the floor-cleaning task can continue. The first useful home robots may win by doing a small part of the benchmark reliably rather than promising the entire room.

What should buyers ask before believing a tidy-up demo?

A living-room demo is worth watching closely, but the questions matter more than the music in the video.

First, ask whether the run is continuous. A robot that completes a ten-step cleanup with no resets, cuts, or human intervention is showing a different level of autonomy from one that succeeds after repeated attempts.

Second, ask whether the room is meaningfully variable. If every object starts in a known place, the demo may be a choreography test. A real benchmark should randomize the pillow, remote, toys, towel, and bin location.

Third, ask what happens when the robot fails. Dropped objects, blocked paths, and uncertain classifications are normal. A useful home robot should recover, ask, or mark the item for human help. Silent failure is not autonomy.

Fourth, ask whether the robot can personalize storage. Tidying is not only "pick up object." It is "put this person's object where this household expects it." That requires memory, permissions, and a privacy model.

Fifth, ask whether there is an available product, price, support model, and safe fallback. The compare tool is helpful here because two robots can both claim manipulation while being completely different purchases: a research platform, a beta home robot, an enterprise RaaS robot, or a shipping vacuum with a small arm.

The honest ranking: who should care now?

If you want a robot that tidies the whole living room today, wait. The strongest public demos are still demos, and the most capable platforms are not ordinary consumer products.

If you are an early adopter, the most interesting category is not "humanoid" by itself. It is home-trained manipulation with a clear support model. That is why robots like NEO and Memo are worth watching: they are explicitly aimed at domestic chores, not just factory transfer. Their challenge is proving repeatability, privacy, and service support in real homes.

If you are a researcher, caregiver, or developer, Stretch 3 and HSR-style platforms remain relevant because they expose the hard parts instead of hiding them: teleoperation, grasping, navigation, data collection, and assistive use.

If you just want less clutter before the robot vacuum runs, a limited-arm cleaner like the Saros Z70 may be more practical than a $20,000 humanoid. It does one slice of tidying, but it is a slice attached to a job people already buy robots to do.

The bottom line

Living-room tidying is the right benchmark because it is ordinary, messy, and hard to fake. It tests whether a home robot can combine perception, movement, manipulation, memory, and judgment in a place that changes every day.

Figure's Helix 02 demo is an important signal, but it should raise the standard for everyone else rather than end the conversation. The next useful home robot does not need to look the most human. It needs to make reliable progress in the room where humans actually live.

For now, treat every tidy-up claim as a benchmark result, not a buying promise. The winning robot will be the one that can clean up a living room twice, in two different homes, with the couch moved, the toys changed, the pillow in a new place, and no engineer waiting just out of frame.