Popular repositories Loading
-
-
-
KernelBench-v2
KernelBench-v2 PublicForked from ScalingIntelligence/KernelBench
KernelBench v2: Can LLMs Write GPU Kernels? - Benchmark with Torch -> Triton (and more!) problems
-
ai-scientist-artefacts-v1
ai-scientist-artefacts-v1 PublicArtefacts from the first complete run of the Lossfunk AI Scientist pipeline for paper accepted at Agents4Science 2025.
-
Repositories
- consciousness-sweep Public
Agent-assisted literature discovery for the functional role of consciousness. Screens titles, abstracts and full texts for studies linking awareness to how information is used. Includes citation expansion, PDF retrieval, resumable runs and source-grounded decision trails, producing prospects for human review.
- Causal-geometric-grokking Public
- Self-Driving-Symbolic-Policy Public
Physics based Symbolic Policy for Self driving in Carla written and acted upon by GPT 6 Astra. A starting point for future work on symbolically grounded world models.
- behaving-rnns Public
Teaching recurrent vision models when to respond. A pretrained BLT network and learned decision controller generate choices and response times without training on human reaction times. Includes behavioral comparisons, model checkpoints, and code to reproduce the experiments and figures.
- authority-bias Public
Language models can abandon correct answers when a "verified" source disagrees. We study how this source deference differs from agreement with a user, and whether activation interventions can reduce it. This repository contains code for behavioral evaluations, attribution patching, and steering experiments.
- scene-memory-tpr Public
Testing whether simple tensor-product representations can explain and selectively edit a recurrent network’s memory for letters and locations.
- crafter-symbolic-solver Public
A pure-Python heuristic agent for Crafter, exploring diamond collection and achievement scores within 10,000 steps. Reads RGB observations, keeps memory and runs on CPU. Includes benchmarks, a live viewer and GIF export.
- verbalizing-eval-awareness Public
A fun experiment to see whether models are self-aware and can actively state that they are being evaluated or not.
People
This organization has no public members. You must be a member to see who’s a part of this organization.
Top languages
Loading…
Most used topics
Loading…