Mastodawn

This research introduces Open-YOLO 3D, a novel method using 2D object detectors for high-speed, open-vocabulary 3D instance segmentation. https://hackernoon.com/a-new-approach-to-3d-scene-understanding-replacing-heavy-segmentation-models-for-a-16x-speedup #zeroshotlearning

A New Approach to 3D Scene Understanding: Replacing Heavy Segmentation Models for a 16x Speedup | HackerNoon

This research introduces Open-YOLO 3D, a novel method using 2D object detectors for high-speed, open-vocabulary 3D instance segmentation.

HackerNoon Aug 23

We tested GPT-4, Llama-2, and more on symbolic puzzles—see why even the strongest LLMs fail without fine-tuning. https://hackernoon.com/why-llms-struggle-with-arithmetic-puzzles #zeroshotlearning

Why LLMs Struggle with Arithmetic Puzzles | HackerNoon

We tested GPT-4, Llama-2, and more on symbolic puzzles—see why even the strongest LLMs fail without fine-tuning.

HackerNoon Aug 23

Can LLMs solve math? This study explores puzzles, synthetic data, and fine-tuning to push AI’s limits in reasoning and extrapolation. https://hackernoon.com/testing-large-language-models-on-math-puzzles #zeroshotlearning

Testing Large Language Models on Math Puzzles | HackerNoon

Can LLMs solve math? This study explores puzzles, synthetic data, and fine-tuning to push AI’s limits in reasoning and extrapolation.

HackerNoon Aug 23

Fine-tuning boosts AI reasoning: models trained on 100M samples achieve higher pass@1 rates in puzzle-solving across in-distribution and OOD tests. https://hackernoon.com/evaluating-fine-tuned-llms-on-reasoning-puzzles #zeroshotlearning

Evaluating Fine-Tuned LLMs on Reasoning Puzzles | HackerNoon

Fine-tuning boosts AI reasoning: models trained on 100M samples achieve higher pass@1 rates in puzzle-solving across in-distribution and OOD tests.

HackerNoon Aug 23

A new dataset turns arithmetic puzzles into a benchmark for AI, testing LLaMA’s reasoning with LoRA fine-tuning and OOD evaluation. https://hackernoon.com/a-framework-for-synthesizing-arithmetical-puzzle-datasets-for-large-language-models #zeroshotlearning

A Framework for Synthesizing Arithmetical Puzzle Datasets for Large Language Models | HackerNoon

A new dataset turns arithmetic puzzles into a benchmark for AI, testing LLaMA’s reasoning with LoRA fine-tuning and OOD evaluation.

HackerNoon Aug 23

Fine-tuning LLMs with synthetic data boosts multi-step mathematical reasoning and improves zero-shot performance on novel benchmarks. https://hackernoon.com/how-llms-learn-to-solve-complex-math #zeroshotlearning

How LLMs Learn to Solve Complex Math | HackerNoon

Fine-tuning LLMs with synthetic data boosts multi-step mathematical reasoning and improves zero-shot performance on novel benchmarks.

DurstewitzLab May 21

Can time series (TS) #FoundationModels (FM) like Chronos zero-shot generalize to unseen #DynamicalSystems (DS)? #AI

No, they cannot!

But *DynaMix* can, the first TS/DS foundation model based on principles of DS reconstruction, capturing the long-term evolution of out-of-domain DS: https://arxiv.org/pdf/2505.13192v1

Unlike TS foundation models, DynaMix exhibits #ZeroShotLearning of long-term stats of unseen DS, incl. attractor geometry & power spectrum, w/o *any* re-training, just from a context signal.
It does so with only 0.1% of the parameters of Chronos & 10x faster inference times than the closest competitor.

It often even outperforms TS FMs on forecasting diverse empirical time series, like weather, traffic, or medical data, typically used to train TS FMs.
This is surprising, cos DynaMix’ training corpus consists *solely* of simulated limit cycles & chaotic systems, no empirical data at all!

And no, it’s neither based on Transformers nor Mamba – it’s a new type of mixture-of-experts architecture based on the recently introduced AL-RNN (https://proceedings.neurips.cc/paper_files/paper/2024/file/40cf27290cc2bd98a428b567ba25075c-Paper-Conference.pdf), specifically trained for DS reconstruction.

Remarkably, DynaMix not only generalizes zero-shot to novel DS, but it can even generalize to new initial conditions and regions of state space not covered by the in-context information.

We dive a bit into the reasons why current time series FMs not trained for DS reconstruction fail, and conclude that a DS perspective on time series forecasting & models may help to advance the #TimeSeriesAnalysis field.

DSN - Data Science Nigeria Jan 31

Hello @dsnai Tech Tribe!

As we end the week, let's close the day with this Trivia.

What does the term "zero-shot learning" mean in the context of machine learning?

Share your insights in the comments - can't wait to see how the DSN community tackles this one!

#TechTrivia #MachineLearning #ZeroShotLearning

Tech Chilli Jun 26, 2024

Innovative Humanoid Robot Alter3 Uses GPT-4 to Execute Detailed Commands.

See here - https://techchilli.com/robotics/innovative-humanoid-robot-alter3-uses-gpt-4-to-execute-detailed-commands/

#GPT4 #Alter3 #HumanoidRobot #AI #Robotics #Innovation #ZeroShotLearning #TechAdvancement #FutureOfAI #AIinteraction #RobotCapabilities #TechRevolution

Innovative Humanoid Robot Alter3 Uses GPT-4 to Execute Detailed Commands

Alter3, a humanoid robot with GPT-4 integration that shows spontaneous motion creation, with no explicit programming required, executes intricate movements like taking selfies, imitating ghosts, and even playing out entire scenarios.

Tech Chill