Program-as-Weights: A Programming Paradigm for Fuzzy Functions
Paper โข 2607.02512 โข Published โข 76
ZeroGPU Explorers
community
AI & ML interests
None defined yet.
Recent Activity
View all activity
yuntian-dengย
submitted a
paper to Daily Papers 2 days ago
eienmojikiย
posted an update 10 days ago
Post
141
Hi everyone,
I've created a Gradio space for embedding and extracting invisible watermarks in images:
๐ eienmojiki/blind-watermark-studio
It supports hiding text, images, and bit arrays using the DWT-DCT-SVD algorithm.
Credits:
- Original library: https://github.com/guofei9987/blind_watermark
- Author: Guo Fei
:).
I've created a Gradio space for embedding and extracting invisible watermarks in images:
๐ eienmojiki/blind-watermark-studio
It supports hiding text, images, and bit arrays using the DWT-DCT-SVD algorithm.
Credits:
- Original library: https://github.com/guofei9987/blind_watermark
- Author: Guo Fei
:).
yuntian-dengย
authored 3
papers 27 days ago
Long Grounded Thoughts: Distilling Compositional Visual Reasoning Chains at Scale
Paper โข 2511.05705 โข Published โข 10
DetectZoo: A Unified Toolkit for AI-Generated Content Detection Across Text, Audio, and Image Modalities
Paper โข 2606.04205 โข Published
Code2LoRA: Hypernetwork-Generated Adapters for Code Language Models under Software Evolution
Paper โข 2606.06492 โข Published โข 95
gagan3012ย
authored a
paper about 1 month ago
gagan3012ย
submitted a
paper to Daily Papers about 1 month ago
tianbaoxiexxxย
authored 6
papers about 1 month ago
OS-MAP: How Far Can Computer-Using Agents Go in Breadth and Depth?
Paper โข 2507.19132 โข Published
Agent Data Protocol: Unifying Datasets for Diverse, Effective Fine-tuning of LLM Agents
Paper โข 2510.24702 โข Published โข 32
OSWorld-MCP: Benchmarking MCP Tool Invocation In Computer-Use Agents
Paper โข 2510.24563 โข Published โข 23
Qwen3-VL Technical Report
Paper โข 2511.21631 โข Published โข 164
RLAnything: Forge Environment, Policy, and Reward Model in Completely Dynamic RL System
Paper โข 2602.02488 โข Published โข 36
CUA-Gym: Scaling Verifiable Training Environments and Tasks for Computer-Use Agents
Paper โข 2605.25624 โข Published โข 34
Locutusqueย
posted an update about 1 month ago
Post
339
๐ Introducing Esmeralda-Llama-3.1-8B-control
The first release in the Esmeralda model family by Locutusque.
This model is intentionally small and experimental โ a control/baseline proof-of-concept designed to answer one question:
ยซโHow strong is my new "Locutusque/esmeralda-agentic" dataset before scaling to larger runs?โยป
Training Details
- Base: Llama 3.1 8B
- Training precision: bf16 mixed precision
- Chat template: modified ChatML
- Dataset size: ~37k examples
- Examples actually used for this run: ~5k
The dataset includes:
- multi-turn agentic traces
- reasoning traces
- structured assistant behavior
- generalist instruction data
Benchmark Results
Compared against:
- Llama 3.1 8B Instruct
- Hermes-3-Llama-3.1-8B
HumanEval
57.3 โ Esmeralda
56.1 โ Llama 3.1 Instruct
52.4 โ Hermes-3
MBPP
53.2 โ Esmeralda
56.8 โ Llama 3.1 Instruct
48.2 โ Hermes-3
GPQA Diamond
15.7 โ Esmeralda
15.7 โ Llama 3.1 Instruct
18.2 โ Hermes-3
EQ-Bench
59.2 โ Esmeralda
61.1 โ Llama 3.1 Instruct
63.1 โ Hermes-3
EQ-Bench Parseable (Syntax Stability)
๐ฅ 100.0% โ Esmeralda
92.4% โ Llama 3.1 Instruct
91.2% โ Hermes-3
Here Be Dragons ๐
I also experimented with a new TruthfulQA free-generation evaluation setup.
- Responses were judged by Gemma 4 26B A4B
- The judge compared generations directly against ground-truth answers
- Models were evaluated in 8-bit quantized form to speed up inference
TruthfulQA (LLM Judge)
0.682 โ Esmeralda-Llama-3.1-8B-control
0.587 โ Hermes-3-Llama-3.1-8B (reported MC2 score; methodology differs)
For a lightweight control run trained on only a fraction of the dataset, Iโm pretty encouraged by the results.
The model is released under the standard Llama 3.1 license, and Iโd genuinely love feedback from people testing it in real workflows.
Model: Locutusque/Esmeralda-Llama-3.1-8B-control
Dataset: Locutusque/esmeralda-agentic
The first release in the Esmeralda model family by Locutusque.
This model is intentionally small and experimental โ a control/baseline proof-of-concept designed to answer one question:
ยซโHow strong is my new "Locutusque/esmeralda-agentic" dataset before scaling to larger runs?โยป
Training Details
- Base: Llama 3.1 8B
- Training precision: bf16 mixed precision
- Chat template: modified ChatML
- Dataset size: ~37k examples
- Examples actually used for this run: ~5k
The dataset includes:
- multi-turn agentic traces
- reasoning traces
- structured assistant behavior
- generalist instruction data
Benchmark Results
Compared against:
- Llama 3.1 8B Instruct
- Hermes-3-Llama-3.1-8B
HumanEval
57.3 โ Esmeralda
56.1 โ Llama 3.1 Instruct
52.4 โ Hermes-3
MBPP
53.2 โ Esmeralda
56.8 โ Llama 3.1 Instruct
48.2 โ Hermes-3
GPQA Diamond
15.7 โ Esmeralda
15.7 โ Llama 3.1 Instruct
18.2 โ Hermes-3
EQ-Bench
59.2 โ Esmeralda
61.1 โ Llama 3.1 Instruct
63.1 โ Hermes-3
EQ-Bench Parseable (Syntax Stability)
๐ฅ 100.0% โ Esmeralda
92.4% โ Llama 3.1 Instruct
91.2% โ Hermes-3
Here Be Dragons ๐
I also experimented with a new TruthfulQA free-generation evaluation setup.
- Responses were judged by Gemma 4 26B A4B
- The judge compared generations directly against ground-truth answers
- Models were evaluated in 8-bit quantized form to speed up inference
TruthfulQA (LLM Judge)
0.682 โ Esmeralda-Llama-3.1-8B-control
0.587 โ Hermes-3-Llama-3.1-8B (reported MC2 score; methodology differs)
For a lightweight control run trained on only a fraction of the dataset, Iโm pretty encouraged by the results.
The model is released under the standard Llama 3.1 license, and Iโd genuinely love feedback from people testing it in real workflows.
Model: Locutusque/Esmeralda-Llama-3.1-8B-control
Dataset: Locutusque/esmeralda-agentic
alvarobarttย
posted an update about 1 month ago
Post
443
Open agents on AWS SageMaker AI with open models from the Hugging Face Hub!
> Deploy an open model from the Hugging Face Hub on SageMaker AI
> Connect the deployed model to Strands Agents
> Add built-in and custom tools for tool calling
> Expose external capabilities through MCP integration
> Bonus: talk to your agent and visualize traces with Gradio
https://alvarobartt.com/agents-on-aws-sagemaker
> Deploy an open model from the Hugging Face Hub on SageMaker AI
> Connect the deployed model to Strands Agents
> Add built-in and custom tools for tool calling
> Expose external capabilities through MCP integration
> Bonus: talk to your agent and visualize traces with Gradio
https://alvarobartt.com/agents-on-aws-sagemaker
alvarobarttย
posted an update about 2 months ago
Post
3346
Latest
TL; DR MoEs can be misleading to reason about from active parameters alone, since each token only activates a subset of experts, while the serving setup still needs to account for the full resident memory footprint.
๐ง
๐๏ธ Dense models usually load and use most weights every forward pass, while MoEs load many experts but only route each token to a few of them
โก Active params isn't the same as memory footprint, especially for sparse architectures
๐ฆ Runtime memory is about what is used per request/token, while loading memory also includes the expert weights that need to be resident
๐ KV cache can still dominate depending on context length, batch size, and concurrency
๐ Expert Parallelism (EP) helps shard experts across accelerators when expert weights dominate
๐ Data Parallelism (DP) + EP is often a good fit for throughput-oriented MoE serving
Check the repository at https://github.com/alvarobartt/hf-mem
hf-mem release added a breakdown of Mixture-of-Experts (MoE) memory usage!TL; DR MoEs can be misleading to reason about from active parameters alone, since each token only activates a subset of experts, while the serving setup still needs to account for the full resident memory footprint.
๐ง
hf-mem now splits MoE memory into base model weights, routed experts, and KV cache๐๏ธ Dense models usually load and use most weights every forward pass, while MoEs load many experts but only route each token to a few of them
โก Active params isn't the same as memory footprint, especially for sparse architectures
๐ฆ Runtime memory is about what is used per request/token, while loading memory also includes the expert weights that need to be resident
๐ KV cache can still dominate depending on context length, batch size, and concurrency
๐ Expert Parallelism (EP) helps shard experts across accelerators when expert weights dominate
๐ Data Parallelism (DP) + EP is often a good fit for throughput-oriented MoE serving
Check the repository at https://github.com/alvarobartt/hf-mem
Post
2765
I'm releasing OpenCS2 a 11TB dataset of around 5000 hours of counter strike gameplay recording.
- HD resolution - 1280ร720 ยท 32 fps
- For each frame keyboard and mouse + world state (player position, velocity, weapon ...)
- HD Stereo audio
- All 10 players perspective
https://huggingface.co/collections/blanchon/opencs2
- HD resolution - 1280ร720 ยท 32 fps
- For each frame keyboard and mouse + world state (player position, velocity, weapon ...)
- HD Stereo audio
- All 10 players perspective
https://huggingface.co/collections/blanchon/opencs2
- 1 reply
seyedhamidrezaย
authored a
paper about 2 months ago
seyedhamidrezaย
submitted a
paper to Daily Papers about 2 months ago
Post
3405
A small model that struggled against a random opponent now beats GPT-5-mini at tic-tac-toe
I took LiquidAI/LFM2-2.6B and trained it through play.
๐งโ๐ณ Here's how:
1๏ธโฃ Build a solid RL env with Verifiers (Prime Intellect)
2๏ธโฃ Generate synthetic data: <200 games sampled from GPT-5-mini playing in the env
3๏ธโฃ SFT warm-up to teach format
4๏ธโฃ Group-based RL (CISPO) against opponents making 20-70% random moves
5๏ธโฃ RL again with stronger opponents (0-25% random moves) + 1.25 temperature to push exploration and shake off suboptimal strategies
Done! Beats GPT-5-mini ๐
---
๐ฎ Play against the model: anakin87/LFM2-2.6B-mr-tictactoe
๐ค Model: anakin87/LFM2-2.6B-mr-tictactoe
๐ Walkthrough/course: https://github.com/anakin87/llm-rl-environments-lil-course
๐ค Dataset and checkpoints: https://huggingface.co/collections/anakin87/lfm2-26b-mr-tic-tac-toe
I took LiquidAI/LFM2-2.6B and trained it through play.
๐งโ๐ณ Here's how:
1๏ธโฃ Build a solid RL env with Verifiers (Prime Intellect)
2๏ธโฃ Generate synthetic data: <200 games sampled from GPT-5-mini playing in the env
3๏ธโฃ SFT warm-up to teach format
4๏ธโฃ Group-based RL (CISPO) against opponents making 20-70% random moves
5๏ธโฃ RL again with stronger opponents (0-25% random moves) + 1.25 temperature to push exploration and shake off suboptimal strategies
Done! Beats GPT-5-mini ๐
---
๐ฎ Play against the model: anakin87/LFM2-2.6B-mr-tictactoe
๐ค Model: anakin87/LFM2-2.6B-mr-tictactoe
๐ Walkthrough/course: https://github.com/anakin87/llm-rl-environments-lil-course
๐ค Dataset and checkpoints: https://huggingface.co/collections/anakin87/lfm2-26b-mr-tic-tac-toe