If your problem can be solved that way, great! It is a Tool That Works. However, the challenges around single environment, limited sample regimes should still call into question our level of understanding of the basic learning process, and deeper understanding there is plausibly
@id_aa_carmack
-
256 Tb/s Fiber Optic Data Rates Demonstrated Over 200 km
By
–
256 Tb/s data rates over 200 km distance have been demonstrated on single mode fiber optic, which works out to 32 GB of data in flight, “stored” in the fiber, with 32 TB/s bandwidth. Neural network inference and training can have deterministic weight reference patterns, so it is
-
#PaperADay January recap: Reading one paper daily challenge
By
–
#PaperADay recap On January 8th, I set out to read and take notes on one paper each weekday for the rest of the month. I missed one day due to a funeral, and another day due to bad time management, but not too bad. I probably averaged a bit over 2 hours on each of them, which
-
DreamerV3: World Models for 150+ Diverse Domains
By
–
#PaperADay 15
2024: Mastering Diverse Domains through World Models
(DreamerV3) https://danijar.com/project/dreamerv3/
… https://arxiv.org/pdf/2301.04104 Applies the latest Dreamer model to over 150 diverse tasks, achieving state of the art scores on many of them, but most notably, applies it to mining -
DreamerV2: Mastering Atari with Discrete World Models
By
–
#PaperADay 14
2022: MASTERING ATARI WITH DISCRETE WORLD MODELS
(DreamerV2) https://danijar.com/project/dreamerv2/
… https://arxiv.org/pdf/2010.02193 DreamerV1 was mostly targeted at continuous control tasks, but it also demonstrated basic playing of Atari games and DMLab tasks. DreamerV2 improved the -
Dream to Control: Learning Behaviors by Latent Imagination
By
–
#PaperADay 13
2020: DREAM TO CONTROL: LEARNING BEHAVIORS BY LATENT IMAGINATION https://danijar.com/project/dreamer/
… More than doubled the performance of PlaNet, and beat the state-of-the-art model-free algorithm of the day that used many more environment steps. PlaNet (#PaperADay 12) wasn't -
Biological Brain Sparsity and GPU Simulation Challenges
By
–
You can argue that bio brains have vastly more weights that are mostly sparse, because the space of neurons that could have been connected to is very large, with synapses exploring and getting pruned. Simulating bio connectivity would be expensive on GPUs! Bio neurons look good
-
LeJEPA: Provable and Scalable Self-Supervised Learning Without Heuristics
By
–
#PaperADay 10
LeJEPA: Provable and Scalable Self-Supervised Learning Without the Heuristics https://arxiv.org/pdf/2511.08544 The comments on #PaperADay 3 recommended this paper as the state of the art JEPA paper, and it does look much better! They acknowledge that much of the prior JEPA -
Flow Model Architecture: Exploring Layer Configurations and Training Efficiency
By
–
Did you try any different configurations of the flow model than 4 layers? I would generally expect a wider 2 layer to train faster, unless there is some character to the flow problem that needs more abstraction.
-
floq: Training Critics via Flow-Matching for Scaling Compute in Value-Based RL
By
–
#PaperADay 9
floq: Training Critics via Flow-Matching for Scaling Compute in Value-Based RL https://arxiv.org/pdf/2509.06863 In theory, value based reinforcement learning is a regression problem, which is most naturally addressed with an MSE loss. However, there are a bunch of subtle