In a Training Loop 🔄

John Smith PRO

John6666

John6666cat

AI & ML interests

None yet

Recent Activity

updated a dataset about 3 hours ago

John6666/forum3

upvoted an article about 5 hours ago

Introducing AI chunking to semchunk

reacted to abdurrahmanbutler's post with 🚀 about 5 hours ago

Isaacus just shipped a major update to semchunk: AI-powered chunking based on a document’s knowledge graph representation⚡ This isn’t a tweak on existing semantic chunking. It’s an entirely new paradigm, built on hierarchical document segmentation rather than heuristics or standard embedding-based semantic approaches. We benchmarked our AI chunking mode across a full RAG pipeline against popular alternatives like LangChain, Chonkie, and our own non-AI semantic chunker. The results were clear: semchunk’s AI mode delivered a 15% relative improvement in RAG correctness over Chonkie. It also produced more aesthetically coherent and readable when judged by a human evaluator while also being faster than all other chunking methods when run on a consumer PC. These gains are powered by Isaacus' Kanon 2 Enricher model, which performs hierarchical document segmentation and directly powers our AI chunking mode. As far as we know, semchunk is one of the first chunking libraries to offer true AI-powered, hierarchical-segmentation-based chunking, and the results show how much better RAG can get when chunking improves. https://huggingface.co/blog/isaacus/introducing-ai-chunking-to-semchunk

View all activity

Organizations

reactedto abdurrahmanbutler's post with 🚀 about 5 hours ago

Post

Isaacus just shipped a major update to semchunk: AI-powered chunking based on a document’s knowledge graph representation⚡

This isn’t a tweak on existing semantic chunking. It’s an entirely new paradigm, built on hierarchical document segmentation rather than heuristics or standard embedding-based semantic approaches.

We benchmarked our AI chunking mode across a full RAG pipeline against popular alternatives like LangChain, Chonkie, and our own non-AI semantic chunker. The results were clear: semchunk’s AI mode delivered a 15% relative improvement in RAG correctness over Chonkie. It also produced more aesthetically coherent and readable when judged by a human evaluator while also being faster than all other chunking methods when run on a consumer PC.

These gains are powered by Isaacus' Kanon 2 Enricher model, which performs hierarchical document segmentation and directly powers our AI chunking mode.

As far as we know, semchunk is one of the first chunking libraries to offer true AI-powered, hierarchical-segmentation-based chunking, and the results show how much better RAG can get when chunking improves.

https://huggingface.co/blog/isaacus/introducing-ai-chunking-to-semchunk

reactedto karstenskyt's post with 👍 about 5 hours ago

Post

🚀 𝗟𝘂𝘅𝘂𝗿𝘆 𝗟𝗮𝗸𝗲𝗵𝗼𝘂𝘀𝗲 𝗨𝗜 𝗨𝗽𝗴𝗿𝗮𝗱𝗲: 𝗠𝗶𝗴𝗿𝗮𝘁𝗶𝗻𝗴 𝘁𝗼 𝗧𝗮𝗶𝗽𝘆

I’m excited to share a major frontend architecture upgrade for the (𝘙𝘪𝘨𝘩𝘵! 𝘓𝘶𝘹𝘶𝘳𝘺!) 𝘓𝘢𝘬𝘦𝘩𝘰𝘶𝘴𝘦 open-source soccer analytics platform. We have officially migrated the dashboard UI from 𝗦𝘁𝗿𝗲𝗮𝗺𝗹𝗶𝘁 to 𝗧𝗮𝗶𝗽𝘆, and it is now live on Hugging Face Spaces.

𝗪𝗵𝘆 𝘁𝗵𝗲 𝘀𝘄𝗶𝘁𝗰𝗵?

While Streamlit was fantastic for prototyping our initial 12 dashboards, we started running into some persistent "jittery" rendering issues as the app grew more complex—specifically when handling 𝟯𝟴𝗠+ 𝘁𝗿𝗮𝗰𝗸𝗶𝗻𝗴 𝗳𝗿𝗮𝗺𝗲𝘀 across 5 professional data providers.

Rebuilding the app in Taipy (running via the Docker SDK on HF Spaces) immediately smoothed out those state-management hiccups. The difference is exceptionally noticeable when interacting with our native Plotly charts, like the Pass Networks and Pitch Control surfaces.

More importantly, this architectural switch sets the foundation for our next major roadmap milestone. Taipy natively excels at managing 𝗮𝘀𝘆𝗻𝗰𝗵𝗿𝗼𝗻𝗼𝘂𝘀 𝗯𝗮𝗰𝗸𝗴𝗿𝗼𝘂𝗻𝗱 𝘁𝗮𝘀𝗸𝘀 and 𝗮𝗱𝘃𝗮𝗻𝗰𝗲𝗱 𝘄𝗼𝗿𝗸𝗳𝗹𝗼𝘄 𝘁𝗿𝗮𝗰𝗸𝗶𝗻𝗴—capabilities we will be heavily leaning into as we start rolling out some advanced, long-running ML training pipelines soon. 👀

Check out the smoother, faster V2 of the platform here:
luxury-lakehouse/soccer-analytics-app

reactedto wassemgtk's post with 🚀 about 5 hours ago

Post

Here is the updated note and benchmark table for your review.

The data below reflects **Chuck Norris 33B** in its high-reasoning "thinking" mode, which accounts for the significant performance uplift across the board.

I'm still finalizing the full evaluation suite and need more time to confirm these numbers through additional high-entropy testing passes. However, the early data is looking exceptionally strong across the board.

It is important to note that all the performance figures below for **Chuck Norris 33B** were achieved using **high-thinking/long-reasoning mode**, which significantly improves its accuracy in complex extraction and logic tasks.
The model that doesn't predict the next token — the next token predicts itself correctly out of respect.

reactedto DedeProGames's post with 🚀 about 5 hours ago

Post

584

Introducing GRM2, a powerful 3 billion parameter model designed for long-term reasoning and high performance in complex tasks.

Even with only 3 billion parameters, it outperforms qwen3-32b in several benchmarks and complex reasoning tasks.

With just 3 billion parameters, it can also generate extensive and complex code with over 1000 lines, utilize tools comparable to larger models, and is perfect for agentic tasks.

GRM2 is licensed under Apache 2.0, making it ideal as a base for FineTune in other tasks.
You can see more here: OrionLLM/GRM2-3b

reactedto datalyes's post with 👀 about 5 hours ago

Post

Hello there ! quick update :

After multiple requests ( very grateful for the interest ) ALL 15 PatenTeb tasks are accessible now ( automatic request approval ).

reactedto juiceb0xc0de's post with 👀 about 5 hours ago

Post

Using GPT-4o as an LLM-as-judge I was able to compare the response capabilities of 4 Llama models which were all trained on the same corpus of a single voice (a bartender's personality) rather than scraped or synthetic data. The models represented are 1B and 3B models with heretic abliterated counterparts. These quotes were pulled from the final judgement after asking the models questions pulled from the following dataset. juiceb0xc0de/chaotic-absurdity

"The potential for even modestly-sized models to possess a semblance of soul isn't just a programmer's fantasy; it's a burgeoning reality."

"Bella-3b, a 3 billion parameter model, startled me with its capacity to latch onto narrative arcs and paint vivid scenes that you'd expect from an imaginative human mind. This model doesn't just compute; it rhapsodizes."

"The bella-1b and the heretic variations served as reminders that attempted creativity can sometimes feel staggeringly off-key. Their attempts often faltered into mechanical, over-processed responses, without the underlying current of emotion or wit."

"Bella-3b's performance not only challenges the preconception that only colossal language models could articulate a 'soul' but also suggests a future where smaller, finely tuned AIs can engage us with character depth and unexpected flair. They may not breathe (not yet), but are we at the dawn of AI characters with true personality? If Bella-3b is any indication, we're at least on the precipice."

https://huggingface.co/collections/juiceb0xc0de/the-bartenders

reactedto salma-remyx's post with ❤️ about 5 hours ago

Post

567

We built an OpenClaw 🦞 skill that sends daily ranked research recommendations to Slack using the Remyx AI CLI.

You define Research Interests (topics, HF models, GitHub repos, blogs etc), Remyx ranks new arXiv papers and repos to find the most relevant resources, and an OpenClaw cron job posts a formatted digest to your team's #research channel every weekday morning.

The tutorial covers the full setup end-to-end: installing the CLI, creating interests, connecting OpenClaw to Slack, installing the Remyx skill, and scheduling the cron. About 15 minutes start to finish.

Tutorial: https://docs.remyx.ai/tutorials/daily-research-digest-slack

Would love to hear how folks are tracking research for their projects. If you give this a try, let us know what you think!

reactedto unmodeled-tyler's post with 🚀 about 5 hours ago

Post

636

Here's a demo of Vessel Browser in action!

Minimax M2.7 was challenged with navigating to a large Ecom site, curate a selection of 5 different products, and add them all to the cart with included reasoning behind choices. (or try it yourself - open source, MIT license, and BYOK!)

npm i @quanta-intellect/vessel-browser

Vessel is a browser that I've been working on which is designed specifically for agents with human-in-the-loop visibility. It comes with a local MCP server allowing any harness that supports custom MCP to control the browser. Additionally, you can BYOK to 8+ different providers (including custom OAI compatible endpoints and local models).

One of my favorite features of the browser is persistent, bi-directional highlighting - meaning that both you AND the agent can highlight anything on the screen and the agent receives the context.

Vessel Browser is unique in that it surfaces available tools contextually to the agent, meaning the agent doesn't have to decide between 80+ tools at any given time, but rather is focused on a subset of tools most applicable to the current state.

Give it a try!

https://github.com/unmodeled-tyler/vessel-browser

1 reply

reactedto kanaria007's post with 🧠 about 5 hours ago

Post

111

✅ Article highlight: *Personal SI-Core* (art-60-047, v0.1)

TL;DR:
What would it mean for a person or household to have an *SI-Core of their own*?

This note sketches *Personal SI-Core / PersonalOS*: a personal-scale runtime where *you remain the primary principal*, goals are explicit, delegations to apps and providers are scoped and revocable, memory is governed, and effectful actions still pass through structured *ID / OBS / MEM / ETH / EVAL / RML* boundaries.

Read:
kanaria007/agi-structural-intelligence-protocols

Why it matters:
• avoids the “50 assistants, 50 KPIs” problem
• keeps platforms from silently optimizing your life for their own goals
• makes delegation visible, capability-scoped, and revocable
• brings consent, auditability, rollout modes, and rollback into everyday life systems

What’s inside:
• *CityOS → PersonalOS* mapping for person/household-scale governance
• personal GoalSurfaces for wellbeing, finances, learning, and schedules
• identity / role / delegation models for apps, employers, banks, and providers
• governed memory across devices, accounts, and external services
• *Personal Genius Traces (PGT)* for replaying good life patterns
• personal *PoLB* modes for safe experiments in routines, budget rules, and automation

Key idea:
Personal SI-Core is not “an AI that runs your life.” It is a way to make personal coordination, delegation, memory, and experimentation *structural, reviewable, and governed*.

reactedto umarbutler's post with 🚀 about 5 hours ago

Post

3140

Isaacus, the AI research company building legal superintelligence, is hiring!

We're looking for passionate engineers who love to build and tinker and want to have an impact on the world. Specifically, we're hiring:
• ML engineers (Australia).
• Data engineers (Australia).
• Full-stack engineers (Australia).
• DevRel engineers (Australia, San Francisco, and London).
• DevOps engineers (Australia, San Francisco, and London).

If you'd like to be a founding employee at one of the few VC-backed LLM research labs in the world, receive generous equity compensation, and work alongside other highly motivated, highly skilled engineers, get in touch: https://isaacus.com/careers

reactedto wassemgtk's post with 👍 1 day ago

Post

113

Releasing Chuck Norris LLM — full SFT fine-tune with chain-of-thought reasoning.

Trained on +100k examples across math, logic, and code. Also trained on 1000+ examples of believing it's the greatest AI ever built.

Its training loss went to zero. The loss function was too afraid to report anything else.

wassemgtk/chuck-norris-llm

reactedto prithivMLmods's post with 🔥🤗 1 day ago

Post

3486

Map-Anything v1 (Universal Feed-Forward Metric 3D Reconstruction) demo is now available on Hugging Face Spaces. Built with Gradio and integrated with Rerun, it performs multi-image and video-based 3D reconstruction, depth, normal map, and interactive measurements.

🤗 Demo: prithivMLmods/Map-Anything-v1
🤗 Model: facebook/map-anything-v1
🤗 Hf-Papers: MapAnything: Universal Feed-Forward Metric 3D Reconstruction (2509.13414)

reactedto aufklarer's post with 🔥 2 days ago

Post

2029

We benchmarked https://github.com/soniqo/speech-swift, our open-source Swift library for on-device speech AI, against Whisper Large v3 (FP16) on LibriSpeech test-clean.

Three models beat it. Two architectural approaches:

Qwen3-ASR (LALM — Qwen3 LLM as ASR decoder, AuT encoder pretrained on ~40M hours) hits 2.35% WER at 1.7B 8-bit, running at 43x real-time on MLX. Greedy decoding matches beam search — the LLM decoder is strong enough that the greedy path is nearly always optimal.

Parakeet TDT (non-autoregressive transducer — FastConformer + TDT joint network) hits 2.74% WER in 634 MB as a CoreML INT8 model on the Neural Engine. No generative hallucination by design. Leaves GPU completely free.

Two findings worth flagging:
- 4-bit quantization is catastrophic for non-English: Korean 6.89% → 19.95% WER on FLEURS. Use 8-bit for multilingual.
- On CoreML, INT8 is 3.3x *faster* than INT4 — opposite of GPU behavior. Native ANE INT8 MACs vs INT4 lookup table indirection.

All numbers reproducible in 15 minutes.

Full article: https://blog.ivan.digital/we-beat-whisper-large-v3-with-a-600m-model-running-entirely-on-your-mac-20e6ce191174

Library: https://github.com/soniqo/speech-swift

Models: Qwen/Qwen3-ASR-0.6B, Qwen/Qwen3-ASR-1.7B, nvidia/parakeet-tdt-0.6b-v2

reactedto cahlen's post with 🚀 2 days ago

Post

2865

It’s wild to me how you can just make shit now.

You can take a weekend with a raspberry pi 5, a pi camera, a 3d printer, and a smidgen of custom fine tuning (wakeword, whisper, tinybert, and pipertts) and you have physical device as a talking personal assistant.

What a time to be alive.

Edge ai, physical ai, ai augmented animatronics… tiny models. Tiny agents.

Going to be a wild year.

5 replies

reactedto DedeProGames's post with 🔥 2 days ago

Post

529

Introducing GRM2, a powerful 3b parameter model designed for long-term reasoning and high performance in complex tasks.

Even with only 3b of parameters, it outperforms qwen3-32b in several benchmarks.

With only 3b of parameters, it can also generate large and complex code of over 1000 lines, use tools in a way comparable to large models, and is perfect for agentic tasks.

GRM2 is licensed under Apache 2.0, making it perfect as a FineTune base for other tasks.

OrionLLM/GRM2-3b

2 replies

reactedto bhalajin's post with 🔥 2 days ago

Post

127

###### CVPR2026 MetaFood Workshop Challenge Alert ######

🍽️ Dishcovery: Mission II VLM Challenge

We’re excited to share that the Dishcovery II Vision-Language Model Challenge is LIVE, as part of the 3rd MetaFood Workshop @ CVPR 2026.

👉 Join the challenge:
https://www.kaggle.com/competitions/dishcovery-mission-ii-cvpr-2026

👉 Dataset collection (Hugging Face):
jesusmolrdv/MTF25-VLM-Challenge-Dataset-Web
jesusmolrdv/MTF25-VLM-Challenge-Dataset-Synth

👉 Workshop details:
https://sites.google.com/view/cvpr-metafood-2026

🔍 Task
Build a Vision-Language Model that aligns food images with text under real-world conditions:
1. Multi-label retrieval → identify relevant ingredients/components
2. Single-label retrieval → select the best dense food description

📦 Dataset Highlights
- 400K+ image–caption pairs
- Mix of real, noisy, and synthetic data
- Designed for fine-grained food understanding
- Reflects real-world multimodal challenges

⚔️ What makes this interesting?
- Not just accuracy → efficiency matters
- Robustness to noise and domain shift
- Fine-grained alignment between visual and semantic concepts
- Benchmark for next-gen VLMs (CLIP, SigLIP, LLaVA-style models)

📅 Timeline (key milestones)
May 1, 2026 → Final predictions + method summary

🚀 Why participate?
- Benchmark your models on large-scale multimodal food data
- Test robustness under realistic conditions
- Gain visibility via a global leaderboard
- Contribute to the growing Food × Vision × Language research space

🍳 The kitchen is heating up — looking forward to seeing what the community builds!

#multimodal #computervision #vlm #deeplearning #datasets #kaggle #huggingface #ai #research

Dataset Citation: Precision at Scale: Domain-Specific Datasets On-Demand (2407.03463)

reactedto Fishtiks's post with ➕ 2 days ago

Post

139

When I was a child, I had a lot of stuffed animals. I say child, but I played with stuffed animals up until I was 15, and only stopped because others said it was weird. I made personalities for them. I could have made "fan art" or something of that nature, but it existed in my imagination, and sometimes, I'd sketch it. I also played with ALICE, which came naturally to me, then.

Well, it turns out that this is all highly autistic stuff, including playing with toys and stories long later than other children. It's also fascinating to me that these are the qualities which, in my opinion, make deeply autistic individuals great clickworkers/trainers in AI. They realize they're curating a personality, partially as an escape from real people and their cruelty, and are okay for that. A lot of autistic will end up needing AI, and that's okay, because it's better to have something and need it than to need it and not have it available. I hope that as AI improves accessibility features, its benefits are considered alongside costs, to provide more functional AI wherever possible, if cheap and energy-efficient enough.

I hope people don't lose their desires to develop their own skills because of AI. I'm not that good of a drawer, and never will be, but I'd hate to see someone just never try because AI is so good. But at the same time, being a ghostwriter, I believe everyone deserves that sort of creative power, and am proud to be involved in bringing it to them. I'm proud to be involved in replacing myself, because I want AI to write better than I do, so one day, you can describe your perfect show, and simply watch it. Some people say that world is horrific. I see it more like when we finally got to stream a large selection of movies rather than just a few cable or satellite selections that were super expensive.

reactedto kostakoff's post with 🔥 2 days ago

Post

126

Just for fun, let's run the Alibaba MNN benchmark on a DGX Spark!

From time to time, I look for something new or unusual in the AI world, and recently I stumbled upon MNN — a direct competitor to llama.cpp.

I found this project intriguing and set a small goal for myself: to run it on my DGX Spark. I was glad to see that MNN is open-source under the Apache 2.0 license, meaning I was free to fork and modify it.

However, MNN had a few issues out of the box:
- No support for CUDA 13.0
- No support for the Blackwell architecture sm_12
- No built-in support for CUDA benchmarking

I tackled these issues one by one and successfully compiled MNN on the DGX Spark. The benchmark results are currently quite low, but at least it works! Patch file here https://github.com/alibaba/MNN/issues/4289#issuecomment-4093931887

Here is the step-by-step guide on how I built MNN:

mkdir mnn && cd mnn
# Get the code
git clone https://github.com/alibaba/MNN.git
cd MNN

# Reset repo to a specific commit
git reset --hard b1d06d68b3366183d157f0703d7b8a8b61ae55b3

# Apply patch for CUDA 13.0
git apply ../my_changes.patch

mkdir build && cd build
# Configure the project
cmake .. \
  -DMNN_CUDA=ON \
  -DMNN_BUILD_LLM=ON \
  -DMNN_SUPPORT_TRANSFORMER_FUSE=ON \
  -DCMAKE_BUILD_TYPE=Release

# Build libraries and executable binaries
cmake --build . --config Release -j$(nproc)
make -j$(nproc)

How to run the test:
- Download the MNN model: taobao-mnn/Qwen3-30B-A3B-MNN
- Run the benchmark:

./MNN/build/llm_bench -m /path/to/qwen/config.json -a cuda -c 2 -p 512 -n 128 -kv true -rep 3

It works!
Hopefully, the MNN developers will add official CUDA 13 support.

llmlaba

reactedto ifaposto's post with 👍 2 days ago

Post

121

I’ve uploaded a minimal, self-contained implementation of manual autograd for a transformer-based classifier in PyTorch. It can help build intuition for what autograd is doing under the hood and is a useful hands-on reference for low-level differentiation in Transformer models, such as writing custom backward passes and tracing how gradients flow through attention blocks.

🐙 GitHub:

https://github.com/ifiaposto/transformer_custom_autograd/tree/main

📓 Colab:

https://colab.research.google.com/drive/1Lt7JDYG44p7YHJ76eRH_8QFOPkkoIwhn

John Smith PRO

AI & ML interests

Recent Activity

Organizations

John6666's activity