I hadn't thought of this in years, but I was recalling the time at my first real job when I tried to create a rudimentary LLM using Dataflex. I was fired before I was able to completely delete it from my PC.
Any day now I expect my wife to get a text reading "CALL ROB"
What if you could supercharge LLM training by dynamically optimizing the data, not just the model?
Researchers from Peking University, Shanghai AI Lab, & the LLaMA-Factory team present DataFlex.
It's a unified framework that smartly selects, mixes, and re-weights training data on the fly during LLM training.
Outperforms standard full-data training on MMLU benchmarks and improves efficiency, offering a reproducible toolkit for data-centric AI.
DataFlex: A Unified Framework for Data-Centric Dynamic Training of Large Language Models
Paper: arxiv.org/abs/2603.26164
Doc: opendcai.github.io/DataFlex-…
Github: github.com/OpenDCAI/DataFlex
Our report: mp.weixin.qq.com/s/QoXfElsL0…
📬 #PapersAccepted by Jiqizhixin
Depends on what you mean. I've programmed in a lot over the last 40 years, but I am pretty rusty with most of them.
68000 assembler, arexx, c, c , c#, objective c, cobol, pascal, perl, visual basic, dataflex, pl/sql, java, Javascript, gdscript, python, solidity
Plus languages like ispf for mainframe menuing, xml and html for markup. and i've used many variants of basic like apple basic, blitz basic, dark basic, vbscript, vba.
But in a practical sense, the languages I use regularly (like in the last year), which i would be generally competent with is a pretty small list:
C#, java, javascript, python, vbscript, pl/sql, xml, html... and I at least read through c and c and do minor integrations with it.
Top AI papers this week on @huggingface 🚀
- CARLA-Air: Fly Drones Inside a CARLA World -- A Unified Infrastructure for Air-Ground Embodied Intelligence
- FIPO: Eliciting Deep Reasoning with Future-KL Influenced Policy Optimization (surpasses DeepSeek-R1-Zero and o1-mini)
- LongCat-Next: Lexicalizing Modalities as Discrete Tokens by Meituan
- ShotStream: Streaming Multi-Shot Video Generation for Interactive Storytelling
- Out of Sight but Not Out of Mind: Hybrid Memory for Dynamic Video World Models
- DataFlex: A Unified Framework for Data-Centric Dynamic Training of Large Language Models
- TAPS: Task Aware Proposal Distributions for Speculative Sampling
- The Latent Space: Foundation, Evolution, Mechanism, Ability, and Outlook
My curated papers from this week
Mathematical methods and human thought in the age of AI
paper: arxiv.org/abs/2603.26524
Proofdoors and Efficiency of CDCL Solvers
paper: arxiv.org/abs/2603.26286
On merge-models
paper: arxiv.org/abs/2603.26570
Short proofs in combinatorics and number theory
paper: arxiv.org/abs/2603.29961
From logπ to π: Taming Divergence in Soft Clipping via Bilateral Decoupled Decay of Probability Gradient Weight
paper: arxiv.org/abs/2603.14389
SKILL0: In-Context Agentic Reinforcement Learning for Skill Internalization
paper: arxiv.org/abs/2604.02268
Loop Control Management in Tightly Coupled Processor Arrays (TCPAs)
paper: arxiv.org/abs/2603.28645
LongCat-AudioDiT
tech report: github.com/meituan-longcat/L…
LongCat-Next: Lexicalizing Modalities as Discrete Tokens
paper: arxiv.org/abs/2603.27538
Kernel-Smith: A Unified Recipe for Evolutionary Kernel Optimization
paper: arxiv.org/abs/2603.28342
Announcing 1-bit Bonsai: The First Commercially Viable 1-bit LLMs
whitepaper: github.com/PrismML-Eng/Bonsa…
blog: prismml.com/news/bonsai-8b
Preference-Aligned LoRA Merging: Preserving Subspace Coverage and Addressing Directional Anisotropy
paper: arxiv.org/abs/2603.26299
Realtime-VLA V2: Learning to Run VLAs Fast, Smooth, and Accurate
paper: arxiv.org/abs/2603.26360
DIAL: Decoupling Intent and Action via Latent World Modeling for End-to-End VLA
paper: arxiv.org/abs/2603.29844
Avoid Routing Polarization for OCS-based GPU Clusters
paper: arxiv.org/abs/2603.28168
Weight Tying Biases Token Embeddings Towards the Output Space
paper: arxiv.org/abs/2603.26663
AgentFixer: From Failure Detection to Fix Recommendations in LLM Agentic Systems
paper: arxiv.org/abs/2603.29848
ChartNet: A Million-Scale, High-Quality Multimodal Dataset for Robust Chart Understanding
paper: arxiv.org/abs/2603.27064
VAREX: A Benchmark for Multi-Modal Structured Extraction from Documents
paper: arxiv.org/abs/2603.15118
PruneFuse: Efficient Data Selection via Weight Pruning and Network Fusion
paper: arxiv.org/abs/2603.26138
DataFlex: A Unified Framework for Data-Centric Dynamic Training of Large Language Models
paper: arxiv.org/abs/2603.26164
Rethinking Language Model Scaling under Transferable Hypersphere Optimization
paper: arxiv.org/abs/2603.28743
A Family of LLMs Liberated from Static Vocabularies
paper: arxiv.org/abs/2603.15953
VideoZeroBench: Probing the Limits of Video MLLMs with Spatio-Temporal Evidence Verification
paper: arxiv.org/abs/2604.01569
Marco DeepResearch: Unlocking Efficient Deep Research Agents via Verification-Centric Design
paper: arxiv.org/abs/2603.28376
daVinci-LLM:Towards the Science of Pretraining
paper: arxiv.org/abs/2603.27164
Sharp Capacity Scaling of Spectral Optimizers in Learning Associative Memory
paper: arxiv.org/abs/2603.26554
Working Notes on Late Interaction Dynamics: Analyzing Targeted Behaviors of Late Interaction Models
paper: arxiv.org/abs/2603.26259
On Strengths and Limitations of Single-Vector Embeddings
paper: arxiv.org/abs/2603.29519
🚀 Train better models by tuning your data — not just your model.
Introducing DataFlex: an open-source framework for dynamic data strategy during training.
Built for complex / multi-domain fine-tuning:
• 🗂️Data selection
• 🔀Data mixing
• ⚖️Sample reweighting
DataFlex
A unified data-centric training framework built on LLaMA-Factory, supporting dynamic sample selection, domain mixture adjustment, and sample reweighting with full DeepSpeed ZeRO-3 compatibility.
Fijate que NO, te cuento que este gobierno esta comprando mas que nunca a los proveedores con precios mas altos, y sobre todo protege a los monopolios del estado, si en mi rubro…busca dataflex en guatecompras, estan despilfarrando en ipads, drones y cuanta tonteria… buscalo!
Are there no old school Novell guys out there? I dabbled with Unix when the stuff I was writing in DBase & Dataflex lived on Unix boxes, but when I moved to Novell NetWare I absolutely loved it. Longest uptime I ever saw was 4 years on a NetWare 3.1 box
Cobol, Pascal, Fortran, Basic, Visual Basic, PDP-11 Assembly, 6800 assembly, 8080 assembly, Z80 assembly, K&R C, ANSI C, C , TK/TCL, Dataflex, SQL, Bash, Csh, Ksh, a little Python, and I remember none of it (well fragments - but nothing that would do me any good).
#OjoAlDAto - La municipalidad de #VillaNueva compró 9 computadoras de escritorio para diferentes direcciones de la municipalidad. Le adjudicaron la compra a Dataflex por Q90 mil.
May God's overflowing peace, hope, and joy fill our hearts and homes this Easter Sunday.
Wishing you and your loved ones a blessed and joyful Easter.
From all of us at DataFlex.
#HappyEaster#EasterWishes#BlessedEaster#DataFlexFamily