Who's Who of AI

Tanishq Mathew Abraham, Ph.D.

270 trust @iscienceluvr · 90,086 followers
Why they matter

Tracked through public AI activity and peer connections inside the directory.

AI signals
6
past 30d
Sources
4
distinct domains
Discusiones
0
past 30d
Latest signal
5d ago
View every signal from Tanishq Mathew Abraham, Ph.D. →
What they're sharing

Articles & links

model: https://t.co/Hh1XvEzaWJ abs: https://t.co/T2YtoxMa34

Nemotron-TwoTower: Diffusion Language Modeling with Pretrained Autoregressive Context arxiv.org
AI Weekly's analysis →
  • NVIDIA's Nemotron-TwoTower splits an LM into a frozen autoregressive context tower and a trainable diffusion denoiser with bidirectional block attention.
  • The system is built on Nemotron-3-Nano-30B-A3B, a 30B hybrid Mamba-Transformer MoE backbone, and trained on roughly 2.1 trillion tokens.
  • The authors report retaining 98.7% of the autoregressive baseline's quality while delivering 2.42x higher wall-clock generation throughput.
Read full analysis →
View on Bluesky · ♥ 0 ↻ 0 ↩ 0 · 3 from the directory shared this · 66d ago

Great work to the Meta AI team! Best part of it is they have open-sourced the code and plan to open-source data too! So you should be able to train your own brain-to-text model, assuming you have your own MEG! 😄 code: https://t.co/XF9z4JCzzq

GitHub - facebookresearch/brain2qwerty: Non-invasive decoding of typed sentences from MEG and EEG brain recordings using a convolutional encoder, transformer, and character-level language model. github.com
AI Weekly's analysis →
  • Meta's FAIR lab released Brain2Qwerty v2, a non-invasive MEG-to-text pipeline reaching an average 61% word accuracy across nine volunteers.
  • The system was trained on roughly 22,000 sentences per participant recorded over 10 hours, with the top participant reaching 78% word accuracy.
  • The original Brain2Qwerty study, run with 35 volunteers, is being published in Nature Neuroscience with a v1 MEG character error rate of 32%.
Read full analysis →
View on Bluesky · ♥ 0 ↻ 0 ↩ 0 · 2 from the directory shared this · 62d ago

abs: https://t.co/S5GD4ecvTf

Autodata: An agentic data scientist to create high quality synthetic data arxiv.org
AI Weekly's analysis →
  • Meta researchers introduce Autodata, a method that casts an AI agent as a data scientist iteratively generating and refining synthetic training data.
  • The practical implementation is called Agentic Self-Instruct, and meta-optimizing the data scientist agent itself produced a larger uplift than static methods.
  • On legal reasoning tasks, a 4B parameter model trained on agent-made data reportedly beat a 397B parameter baseline.
Read full analysis →
View on Bluesky · ♥ 0 ↻ 0 ↩ 0 · 2 from the directory shared this · 67d ago

Looks like they have released their technical report. "Prime Agent: A Self-Improving RLM Harness" https://t.co/Lvh3FLuYMg The paper discusses a bit more about the design philosophy behind Prime Agent, particularly its role as an agent harness for long-horizon evaluation. https…

Prime Agent: A Self-Improving RLM Harness arxiv.org
AI Weekly's analysis →
  • Prime Agent, an open-source harness from Prime Intellect, raises ARC-AGI-3 RHAE Best@1 from 30% to 95.5%, per the arXiv abstract posted 24 August 2026.
  • The design pairs a persistent IPython REPL following the Recursive Language Model abstraction with a Continual Harness that preserves histories, memories, skills, prompts and subagent specifications across trajectories.
  • The paper reports parity or better than native and popular harnesses on long-context coding, GPU-kernel generation, emulator construction and autonomous nanoGPT speedruns.
Read full analysis →
View on Bluesky · ♥ 0 ↻ 0 ↩ 0 · 2 from the directory shared this · 5d ago