Pinned Tweet
Is "judgement" more than a socially acceptable cover for incomplete instrumentation?
Like it's okay to ship code w/o evidence now that AI drafts more of the features?
As if engineering is more limited by epistemic ambiguity than process failure?
Like it's okay to ship code w/o evidence now that AI drafts more of the features?
As if engineering is more limited by epistemic ambiguity than process failure?
0
0
0
90
Smells Like ML retweeted
Super-Tuning uses calibration-free magnitude scoring to allocate a sparse tuning budget. It also allows to add LoRA on top of this sparse update. It is a good fit for exploring sparsity alongside higher-rank adaptation. Also contributed by @smellslikeml.
1
1
3
80
Smells Like ML retweeted
@smellslikeml added the Riemannian-preconditioned LoRA optimizer. It is a new LoRA-specific optimizer similar to LoRA-FA or LoRA+ that improves feature learning and does not require tuning of learning rates for LoRA's A and B matrices separately.
1
1
3
111
@jaeh0ng_yoon similarity to the reference implementation is a simple evaluation criteria
but in practice, I'll trade off fidelity to the source method in order to respect constraints of the target repo
A maintainer's merge decision or application-specific benchmark is more relevant here
but in practice, I'll trade off fidelity to the source method in order to respect constraints of the target repo
A maintainer's merge decision or application-specific benchmark is more relevant here
1
0
0
27
with codegen, research repos are cheaper to make than ever before
one way we could repurpose the paper2code candidate pool to evaluate something closer to the contributor/practitioner's pain points upstreaming a method into application
gist.github.com/smellslikeml/d…
one way we could repurpose the paper2code candidate pool to evaluate something closer to the contributor/practitioner's pain points upstreaming a method into application
gist.github.com/smellslikeml/d…
0
0
0
29
Got it exactly reversed!
I don't need a bunch of agents working on one problem someone else framed.
I need one agent focused on addressing the problems in my context.
I don't need a bunch of agents working on one problem someone else framed.
I need one agent focused on addressing the problems in my context.
Introducing autoresearch for arXiv papers: replicate and experiment on any arXiv paper using an army of Claude/Codex agents
For post-training, your agents can use @tinkerapi to launch concurrent RL runs while you watch your experiment tree grow
Try it out now on OpenResearch: openresearch.sh
1
0
1
259
Get Outrider: github.com/remyxai/outrid…
The coding harness for continuous contributions to your repo
The coding harness for continuous contributions to your repo
0
0
0
55
Implementing arxiv 2608.14003 in SGLang
github.com/sgl-project/sg…
Uses an adaptive mask under a captured graph (topology static = k-wide GEMM; mask change = between-replay buffer update via version-gated post_fill; prune steps replay, explore steps eager)
github.com/sgl-project/sg…
Uses an adaptive mask under a captured graph (topology static = k-wide GEMM; mask change = between-replay buffer update via version-gated post_fill; prune steps replay, explore steps eager)
0
0
0
106
IssueTrojanBench stress-tested coding agents with malicious issues: 66.5% bypassed every guardrail in 4,176 runs. Vulnerability: GPT-5.3 Codex 84.8%, GPT-5.4 73.6%, Sonnet 4.6 41.1%.
Agent frameworks accounted for 0% of rejections. #AI #Coding #AISafety
arxiv.org/abs/2607.20759
1
0
5
2,819
TransformerLens recently added support for Qwen2.5-VL base models
Probe and compare SpaceQwen, SpaceThinker attention entropy, residual stream, and logits
colab.research.google.com/drive/14p2l_RJ…
Probe and compare SpaceQwen, SpaceThinker attention entropy, residual stream, and logits
colab.research.google.com/drive/14p2l_RJ…
1
0
0
140
Used this to target a subset of layers for finetuning a qwen3-vl-2B-thinking model on a single A100
colab.research.google.com/drive/1Zk04lDM…
colab.research.google.com/drive/1Zk04lDM…
0
0
0
41
fine-tune w/ unsloth, axolotl, ms-swift, LLaMA-Factory?
each relies on PEFT
so I'm adding a new method
github.com/huggingface/pe…
each relies on PEFT
so I'm adding a new method
github.com/huggingface/pe…
0
0
2
153