When your 7-day #Llama fine-tune crashes at 90%

Async #evals, plots not loading & dating apps?!

Every node logs separately… now what? #DistributedTraining

Not enough #gpu time: academic struggles

The #foundationmodel pre-training trick that improved audio fidelity #ICML

The reward hacking trap in RLHF #foundationmodels #ICML

100 failed runs: debugging gradients in #foundationmodels #ICML

The classification #foundationmodel that refused to learn #ICML

When alpha breaks #diffusionmodels training #ICML

Debugging #foundationmodels: when gradients diverge #ICML

#Foundationmodels fail: cell prediction gone wrong #ICML

From 15 hours to 5: fixing #foundationmodels fine-tuning efficiency #ICML

The code logic bug that stalled ECG #foundationmodels #ICML

Training brainwave #foundationmodel: what went wrong #ICML

When GPU utilization drops in #RL #foundationmodels training #ICML

2-week #debugging of a graph #foundationmodel #ICML

How a gradient bug wasted 32 H100s for 3 weeks in #foundationmodel training #ICML

Debugging spiking loss in protein #foundationmodels #ICML

Fixing #foundationmodels training stability #ICML

Why #reinforcementlearning training timed out at Amazon #AGI #ICML

Leveraging machine-unlearning #neurips

The multitask #RL bug that broke #foundationmodels training #ICML

When the wrong prompt wasted a week of #reinforcementlearning training #ICML

The #foundationmodels tokenization trick for time series #ICML

Evaluating vision-language model sensitivity #neurips

The #MoE #loadbalancing bug that wasted 50K A100 hours #ICML

Debugging fairness in tabular #foundationmodels #ICML

Forking runs: the fix for preempted #foundationmodel training #ICML

Infecting #llm based multiagents #neurips

Debugging misaligned EEG–text #foundationmodels #ICML

The cheating bug in robot learning #foundationmodels #ICML

When causal reasoning #foundationmodels drop 40% #ICML

512 #GPU days lost: FP8 failures in #foundationmodels training #ICML

Adaptive LoRA merging #neurips

Evaluation of #multimodal #llms in medical VQA #neurips

Red-teaming #llms #neurips

Realistic threat model for #llm jailbreaks #neurips

Asynchronous parallel SGD #neurips

Memory-efficient #neuralnetworks training #neurips

Instant transformer adaption via HyperLoRA #neurips

Self-Stitching: efficient transfer learning #neurips

Text-diffusion red-teaming of #llms #neurips

Impossible #generativeai questions with Karol Lasocki #neurips

Large-scale multi-task dataset for in-context RL #neurips

Impossible #genai questions with Yinan Wang #neurips

Adversarial negotiation in generative #llms #neurips

Optimizing attention #neurips

Impossible #generativeai questions with Finn Schmidt #neurips