DATA

→ View original post on X — @ylecun

SandboxAQ Releases Awesome New Dataset for AI Research

By

@ylecun

–

20 June 2025 16h53

Awesome new dataset from @SandboxAQ https://t.co/PtmSmM5W6H
— Yann LeCun (@ylecun) 20 juin 2025

Awesome new dataset from @SandboxAQ

Embeddings Recovery with Minimal Data Requirements

By

–

20 June 2025 14h53

there are some other nice properties: – we can get good performance with very little data
– we recover embeddings that "look" like the original dataset
– when some percentage of the original data is available, we generally select that data first

SELECT: Training Models with Limited Data Access

By

–

20 June 2025 14h53

…but it works we try a bunch of baselines like selecting data based on likelihood or even the top-k per-class our method (SELECT) can train a model to similar performance, with access only to model weights and some web data

Finetuning AI Models: Gradient-Based Data Selection Strategy

By

–

20 June 2025 14h53

to do this, you need TWO sets of model weights: the initial model and a finetune this is realistic. open-weights models often come with two checkpoints instead of one-shot generating data from weights, we select data from the web with gradients that point along the model diff

Reverse Engineering Training Data From Language Model Weights

By

–

20 June 2025 14h53

NEW RESEARCH: Approximating Language Model Training Data from Weights ever wonder how much information is available in an open-weights model? DeepSeek R1 weights are 1.2 TB… what can we learn from all those bits? our method reverses LLM finetuning to recover data:

→ Voir le post original sur X — @flashtweet

Talkwalker monitoring limitations X API access issues

By

@flashtweet

–

20 June 2025 13h11

Ce qui me semble plus problématique, au delà de la notion de souveraineté, c’est que @Talkwalker ne monitore pas X et ne paye pas l’accès à l’API de X. C’était le cas la dernière fois que j’ai voulu consulter l’outil. @TalkwalkerFr
, vous confirmez que c est le cas ? Comment

→ View original post on X — @godofprompt

Combining Market Data with LLM Analysis in 2025

By

@godofprompt

–

20 June 2025 10h15

If you are not combining market data with LLM analysis in 2025, you're behind. Grok 3 isn’t perfect, but it’s good enough to augment decision making in real time.