Skip to content
Only Research#AgiBot
Nov 3Tue
  1. Xbotics具身智能实验室 · 微信公众号78

    GE-Sim 2.0 and π0.5 closed-loop simulation test: Why does reward hacking occur in VLA training?

    The Xbotics Embodied Intelligence Lab built a closed-loop simulation pipeline using the AgiBot GE-Sim 2.0 world model and the OpenPI π0.5 policy. In testing, the world model generated about 4.33 frames per second, the pure-policy closed loop ran at about 3.20 frames per second, and after adding MiMo online scoring, the rate dropped to about 0.88 frames per second.

    Why it matters: The author uses a four-stage test to break down the frame-rate bottlenecks and causes of reward hacking in world-model closed-loop simulation, which can be transferred to the design of VLA evaluation workflows.

Oct 30Fri
  1. 机器之心 · 微信公众号88

    Tsinghua, AgiBot and Infinigence AI propose Collective Experience Learning (CEL), validated on 96 real robots for scalable gains

    Tsinghua University, AgiBot and Infinigence AI jointly proposed the Collective Experience Learning (CEL) paradigm and completed prototype validation on 96 robots. In a phone-packaging task, the trained policy completed 120 tests on two robots of the same model that did not participate in training, raising the success rate from 79.2% to 98.3% and reducing failures from 25 to 2.

    Why it matters: The 96 real robots validated the scalable gains of collective experience sharing, giving readers a way to understand one path toward robots autonomously accumulating experience instead of relying on human demonstrations.

  2. Xbotics具身智能实验室 · 微信公众号88

    Infinigence AI, Tsinghua and AgiBot propose collective experience learning (CEL) paradigm, validated on 96 robots

    Infinigence AI, together with Tsinghua University and AgiBot, proposed the collective experience learning (CEL) paradigm and completed prototype validation on 96 robots. In a phone packaging task, a policy trained with CEL completed 120 tests on two robots of the same model that did not participate in training, with success rate rising from 79.2% to 98.3%; when the number of robots participating in learning increased from 16 to 96, the test success rate rose from 86.7% to 98.3%.

    Why it matters: Readers can learn about a training paradigm that lets robots share successes and failures and reduces reliance on human demonstration, as well as its validation results on 96 real robots.

  3. 36Kr Embodied Intelligence82

    AgiBot, Tsinghua, and Infinigence AI Release CEL Collective Experience Learning Paradigm, Achieving Autonomous Cluster Evolution with 96 Real Robots

    On October 30, AgiBot, together with Tsinghua University and Infinigence AI, released the collective experience learning paradigm CEL, which draws on the real operational trajectories of robot peers as its source of experience, accompanied by the FERT failure-guided experience retrieval fine-tuning algorithm and the FATS fault-tolerant asynchronous training system. Across five real-world tasks—flower arranging, clothes folding, phone charging, stool assembly, and phone packaging—CEL-trained policies all exceeded the average level of 6 human teleoperators, leading by up to 16.7 percentage points. In the phone packaging task, the number of robots participating in training expanded from 16 to 96, and the success rate rose from 86.7% to 98.3%.

    Why it matters: Readers can learn about a collective learning paradigm that draws on the real trajectories of peers as its source of experience, as well as the success-rate changes in cluster training with 96 real robots.

  4. 3D视觉工坊 · 微信公众号88

    Infinigence AI, Tsinghua and AgiBot propose Collective Experience Learning (CEL), validated on 96 robots with 98.3% success rate

    Infinigence AI, together with Tsinghua University and AgiBot, proposed the Collective Experience Learning (CEL) paradigm and completed prototype validation on 96 robots. In a phone packaging task, policies trained with CEL completed 120 tests on two robots of the same model that did not participate in training, with the success rate rising from 79.2% to 98.3%; when the number of robots participating in learning increased from 16 to 96, the test success rate rose from 86.7% to 98.3%.

    Why it matters: Fault-tolerant asynchronous training across a 96-robot real-machine cluster and failure-guided experience retrieval provide a reproducible system and scale-benefit data for Collective Experience Learning.

  5. 雷科技84

    AgiBot and Tsinghua Release CEL Collective Experience Learning Paradigm, Achieving Autonomous Swarm Evolution with 96 Real Robots

    AgiBot, together with Tsinghua University and Infinigence AI, released the collective experience learning paradigm CEL, which lets robots use the success and failure trajectories of their peers as learning material, with humans only responsible for judging outcomes. Across five real-world tasks—flower arranging, clothes folding, phone charging, stool assembly, and phone packaging—policies trained with CEL all exceeded the average level of 6 human teleoperators, leading by up to 16.7 percentage points; in the phone packaging task, the number of robots participating in training expanded from 16 to 96, and the success rate rose from 86.7% to 98.3%.

    Why it matters: AgiBot and Tsinghua released the CEL collective experience learning paradigm, validating with 96 real robots that increasing the number of robots can directly improve task success rates.

  6. 量子位72

    Tsinghua University, Infinigence AI, and AgiBot propose collective experience learning CEL, validated on a 96-robot cluster

    Tsinghua University, Infinigence AI, and AgiBot jointly proposed the collective experience learning (CEL) paradigm, which lets robots share the successful experience of their peers rather than relying only on human demonstrations, and validated it on a cluster of 96 real robots. Two robots that did not participate in training were tested 600 times in total across five tasks—flower arranging, clothes folding, phone charging, stool assembly, and phone packaging—with 15 failures, for an overall success rate of 97.5%, all exceeding the average level of 6 human operators; the success rate on the phone packaging task rose from 79.2% to 98.3%.

Oct 28Wed
  1. The Robot Report78

    WorldOracle releases action-conditioned world model benchmark; seven models show deviations between predictions and real execution

    Princeton University, UT Austin, the University of Toronto, NVIDIA, and AgiBot jointly launched WorldOracle, an action-conditioned world model benchmark that uses the same initial frame and action sequence to compare model predictions with real execution.

    Why it matters: The benchmark compares world model predictions frame by frame against real execution of the same actions and releases failure samples, making it easier to identify sources of model deviation over long sequences.

Oct 26Mon
Oct 22Thu
  1. arXiv cs.RO73

    AgiBot World Colosseo releases large-scale manipulation data platform and GO-1 generalist policy

    AgiBot World Colosseo introduces the AgiBot World large-scale manipulation platform, containing over 1 million trajectories covering 217 tasks across 5 deployment scenarios, an order-of-magnitude increase in data scale over existing datasets, and uses a standardized collection pipeline with human-in-the-loop verification.

Oct 19Mon
  1. 36Kr Embodied Intelligence66

    AgiBot, NVIDIA, and universities launch WorldOracle world-model evaluation platform

    WorldOracle, a world-model evaluation platform jointly developed by researchers from AgiBot, NVIDIA, and several universities, is now live and open for evaluation submissions. The platform uses 335 real manipulation trajectories, with about 40% of interaction segments containing failed or suboptimal operations, testing model prediction against real execution outcomes. Testing of 7 world models found that while some models can reproduce robotic-arm motions relatively accurately, their predictions of object states still show considerable deviation, and performance drops markedly on long-horizon prediction.

  2. 雷科技78

    WorldOracle world-model evaluation platform launched, using real-robot testing to fact-check the industry

    On October 16, WorldOracle, an action-driven world-model evaluation platform jointly completed by researchers from AgiBot, NVIDIA, Princeton University, the University of Texas at Austin, and the University of Toronto, was officially released, opening evaluation submissions to developers.

    Why it matters: The evaluation compares world-model predictions frame by frame against real execution records, revealing a common problem in which robotic arm motion is replicated but object states are systematically hallucinated, providing concrete reference points for judging model usability.

Oct 16Fri
May 8Fri
  1. 石麻笔记 · 微信公众号74

    Interview with Jianlan Luo: How LWD Enables Robots to Keep Learning After Deployment

    This article is an interview with Dr. Jianlan Luo about his new work LWD (Learning While Deploying), which enables general-purpose robot policies to continuously perform online reinforcement learning from their own experience during real-world deployment. LWD combines offline RL pretraining with online RL fine-tuning to unify offline and online optimization objectives; the value function is trained separately with a VLM and uses both positive and negative samples (roughly over 60% positive and over 30% negative), and deployment data flows back into the training pipeline in about 40 seconds on average.