Consciousness workshop at the AGI conference in SF (27-30 July)! If you'd like to contribute, please submit talk proposals: https://
docs.google.com/forms/d/e/1FAI
pQLSer6WjlXvWaINbKBSc5qLiHDZIwpiB_sV-7Uf6STS6Y-s179Q/viewform?usp=dialog
…
RESEARCH
-
Consciousness workshop at AGI conference: call for talk proposals
By
–
-
AI model jump felt like Opus 4 to 4.5, GPT-3.5 to 4
By
–
The jump felt to me like Opus 4 to Opus 4.5, or GPT-3.5 to GPT-4
-
GoalOS Mission OS: The Proof OS for Autonomous AI Work
By
–
AI does not need to produce more output. It needs to produce proof. I’m sharing my paper: GoalOS Mission OS: The Proof OS for Autonomous AI Work The thesis is simple: AI creates output.
GoalOS creates proof. Today’s models can answer.
Agents can act.
But institutions need -
Stanford scholars: AI coding agents lack collaboration, need social intelligence
By
–
AI coding agents lack a critical skill: collaboration. @Stanford scholars believe it’s a solvable problem if we train models in social intelligence.
-
GoalOS Mission OS: The Proof OS for Autonomous AI Work
By
–
AI does not need to produce more output. It needs to produce proof. I’m sharing my paper: GoalOS Mission OS: The Proof OS for Autonomous AI Work The thesis is simple: AI creates output.
GoalOS creates proof. Today’s models can answer.
Agents can act.
But institutions need -
SAM 1/2/3 pods discussion: concepts and intent in videogen
By
–
see our SAM 1/2/3 pods with @nikhilaravi and @josephofiowa
, the rise of concepts is definitely part of it, although imo "intent" is more of a videogen problem (see our @EthanHe_42 pod) than pure CV -
Questioning the halt of new AI model releases from all labs
By
–
So are we just going to stop releasing new models from all labs? How can any lab make this guarantee for any future model. Seems absurd.
-
Apodex 1.0-H: New Benchmark in Deep Research
By
–
Apodex 1.0-H is reported as the new state-of-the-art across both open and closed systems on the public deep-research suite, achieving scores of 90.3 on BrowseComp, 94.4 on DeepSearchQA, and 87.4 on FrontierScience-Olympiad. The weights are released under the Apache 2.0 license: Apodex 1.0-mini at 35B-A3B.
-
Distribution of thought paths in a continuous latent space
By
–
Latent Thought Flow This article moves reasoning into a continuous latent space, but instead of learning a single hidden thought path, it learns a distribution over many paths. Using a continuous GFlowNet, Latent Thought Flow assigns a
-
ExpRL uses reference solutions as reward scaffolds for exploratory RL
By
–
“ExpRL: Exploratory RL for LLM Mid-Training” Sparse reward RL works only when the base model can already find useful reasoning paths, but on hard problems it often gets no signal. This paper uses reference solutions as reward scaffolds instead of imitation targets, letting an