Dominique P... @DominiqueCAPa... · 6h
Two months of ablating π0.5 finetunes on a real manufacturing task. The policy now succeeds 98% of the time.
We're publishing all results, all the data, and every run, including the ones that went nowhere.
What we learned:
1/ 5x more data took us from 63% to 76%. Pure scaling was the weakest lever we found.
2/ Diversity got more out of the same hours. 4h spread across five scenes beat 4h in the eval scene by 30pp.
3/ Quality did even better. One clean, slower hour on top of the 21h: 76% → 90%. That single hour added more than the 17 hours before it.
4/ So we tried quality on its own: 1h of clean data plus 240 rollouts with human interventions. 1.7h in total, 28% → 88%. That's 12 points above the 21h model with a twelfth of the data.
5/ Inference settings add more on top, with no retraining. They took the 21h model from 76% to 93%, and the 21h+1h model from 90% to 98%.
Just as useful: what didn't matter. Learning rate, batch size, relative vs. absolute joints, image augmentation. All less important than we expected.
Full write-up linked below.
[attached image/video at bottom, cut off: robot arms with grippers over a tray of small parts in an aluminium-frame workcell]
Note from Claude Sonnet 5
X post from Dominique P... reporting two months of ablations on pi0.5 finetunes for a real manufacturing task (98% success), with five numbered findings on data scaling, diversity, quality, and inference settings. The embedded media at the bottom is a photo of robot arms at a workcell, cut off by the screenshot edge.
roboticspi0.5finetuningdata qualitydata scalingmanufacturingablations
Jasper Lu @lu__jasper · 3h
This is notable. DeepSeek, a lab usually first to pioneer novel algorithms and architectures, is saying that at this point, the ROI of improving data quality far exceeds that of working on novel post-training algorithms.
I think this has already been true for some time for non-lab practitioners. If you're doing llm post-training, 80% of your effort should go into looking at your data.
This means:
- Hiring experts to dig through your RL tasks
- Sifting through rollouts and sft data by hand to remove suspicious samples. Make sure all tasks are actually passable.
- Making sure your data is diverse in both difficulty and category.
[Embedded image of paper excerpt]
5.1. Post-Training Pipeline
In this release, we refrain from introducing novel post-training algorithms. The overall recipe follows the standard paradigm of supervised fine-tuning (SFT) followed by reinforcement learning (RL) and on-policy distillation (OPD; Gu et al., 2024; Lu and Lab, 2025), without algorithmic modifications beyond well-established practices. Instead, our efforts are concentrated almost entirely on what the model is trained on rather than how it is optimized: we invest in large-scale, automated pipelines for data synthesis and environment construction. Concretely, the pipeline (i) synthesizes diverse, verifiable training tasks together with their reference solutions and reward signals, (ii) procedurally constructs and scales interactive agent environments in which trajectories can be collected and evaluated at low cost, and (iii) applies rigorous filtering, deduplication, and difficulty calibration to ensure data quality and curriculum balance. [highlighted in yellow from here:] We find that, under a fixed and unremarkable optimization procedure, systematic improvements in the scale, diversity, and verifiability of synthesized data and environments account for essentially all of the observed gains. This observation echoes a broader lesson: at the current stage, the marginal return of engineering the data and environment pipeline substantially exceeds that of algorithmic novelty in post-training. [end highlight]
[Quoted tweet]
DeepSeek @deepseek_ai · 13h
[small thumbnail of a bar chart]
🚀 Introducing DeepSeek-V4.1-Flash: smarter, faster, more efficient.
🔹 Introducing the smallest model in our new architecture family, with ...
Note from Claude Sonnet 5
Tweet by Jasper Lu on a DeepSeek-V4.1-Flash technical report passage saying data and environment engineering, not novel post-training algorithms, accounts for the gains, with practical advice on looking at your data. Embeds a highlighted paper excerpt and quotes the DeepSeek announcement tweet.
deepseekpost-trainingdata qualityrl environmentsdeepseek-v4.1
Sichu Lu @lu_sichu · 13h
the biggest issue is that a lot of raw human logged data is just BAD and WRONG i keep saying this because while data cleaning is probably happening at some stages in the training it can not remove raw data issues because human are bad at LOGGING and MEASURING the world around us
[quoted tweet]
Ranjit Jhala @RanjitJhala · Aug 7
TIL a really neat paper by @ShriramKMurthi and Matthew Flatt on error messages in the AI age
arxiv.org/abs/2606.01522
...[cut off]
Note from Claude Sonnet 5
Sichu Lu argues that a core issue with training data is that raw human-logged data is often bad/wrong because humans are bad at logging and measuring the world, and data cleaning can't fully fix this; quote-tweeting Ranjit Jhala's mention of a paper on error messages in the AI age (arxiv.org/abs/2606.01522).
training datadata qualitytwitter
