updated the document. Thanks
Javad Taghia | PhD PRO
javadtaghia
AI & ML interests
Work with agentic systems, agent memory, model fine tuning and unlearning, generative AI, and the development and deployment of open source AI
Recent Activity
liked a Space about 20 hours ago
javadtaghia/ai-written-rank updated a Space about 21 hours ago
javadtaghia/ai-written-rank published a Space about 21 hours ago
javadtaghia/ai-written-rankOrganizations
replied to their post about 22 hours ago
posted an update about 23 hours ago
Post
66
https://huggingface.co/blog/javadtaghia/atmem-239-shared-memory-for-coding-agents-and-what
Review of AtMem Code memory. It worths trying
Review of AtMem Code memory. It worths trying
replied to their post 3 days ago
thinking about your comment, it's fair.
posted an update 4 days ago
Post
131
It is not always positive results that worth noting.
I ran a long run of training to improve AtMem (agent memory) on storage path using Laya as a decision model. However, I found it did not give me any better result than original AtMem.
Indeed we can use laya on the retrieval path, we have used than by pluging Jev. So we kinda proved something that is still a gap in memory systems (improving storage) is not as easy as it sounds. Check details https://huggingface.co/blog/javadtaghia/we-fine-tuned-a-memory-classifier-for-our-agent-he
I ran a long run of training to improve AtMem (agent memory) on storage path using Laya as a decision model. However, I found it did not give me any better result than original AtMem.
Indeed we can use laya on the retrieval path, we have used than by pluging Jev. So we kinda proved something that is still a gap in memory systems (improving storage) is not as easy as it sounds. Check details https://huggingface.co/blog/javadtaghia/we-fine-tuned-a-memory-classifier-for-our-agent-he
- 4 replies
posted an update 11 days ago
Post
68
AtMem 2.3.8 beat our recorded local Mem0 and AgentRunbook-R configurations on the LongMemEval-V2 development sample: 5/23 vs 1/23 and 4/23.
On DolphinBench, AtMem scored 9/30 vs Mem0’s 7/30.
Please check
https://huggingface.co/blog/javadtaghia/atmem-2-3-8
https://github.com/aetna000/atmem/releases/tag/v2.3.8
On DolphinBench, AtMem scored 9/30 vs Mem0’s 7/30.
Please check
https://huggingface.co/blog/javadtaghia/atmem-2-3-8
https://github.com/aetna000/atmem/releases/tag/v2.3.8
posted an update 13 days ago
Post
113
RLCD: Are We Moving from AI Models That Generate Answers to Models That Make Calibrated Decisions?
https://huggingface.co/blog/javadtaghia/rlcd-are-we-moving-from-ai-models-that-generate-an
https://huggingface.co/blog/javadtaghia/rlcd-are-we-moving-from-ai-models-that-generate-an
posted an update 22 days ago
Post
126
https://huggingface.co/blog/javadtaghia/the-100-frontier-compute-problem
What happens when AI stops answering questions and starts working for hours?
I have been thinking about the economics of frontier coding models.
What happens when AI stops answering questions and starts working for hours?
I have been thinking about the economics of frontier coding models.
posted an update 23 days ago
Post
117
I’m preparing to submit my article to arXiv and need an endorsement for the cs.AI category.
If you’re eligible to endorse submissions in this area and would be willing to help, I’d really appreciate it. Please DM me and I can send you the arXiv endorsement link.
Thank you!
from arXiv cs.AI community, can anyone support me?
If you’re eligible to endorse submissions in this area and would be willing to help, I’d really appreciate it. Please DM me and I can send you the arXiv endorsement link.
Thank you!
from arXiv cs.AI community, can anyone support me?
reacted to their post with 🚀 3 months ago
Post
230
We have published a complete paper about AtMem.Ai
Check it out here.
Please refer to the technical paper below.
https://github.com/aetna000/aetnamem/blob/e24d5aa74e62dbebeeb28a2e49b5c4d2006bbb4d/paper/aetnamem-control-plane.pdf
Check the benchmark
https://huggingface.co/spaces/Aetna000/MemoryStackBench
Check it out here.
Please refer to the technical paper below.
https://github.com/aetna000/aetnamem/blob/e24d5aa74e62dbebeeb28a2e49b5c4d2006bbb4d/paper/aetnamem-control-plane.pdf
Check the benchmark
https://huggingface.co/spaces/Aetna000/MemoryStackBench
- 4 replies
replied to their post 7 months ago
This comment has been hidden
posted an update 9 months ago
Post
230
We have published a complete paper about AtMem.Ai
Check it out here.
Please refer to the technical paper below.
https://github.com/aetna000/aetnamem/blob/e24d5aa74e62dbebeeb28a2e49b5c4d2006bbb4d/paper/aetnamem-control-plane.pdf
Check the benchmark
https://huggingface.co/spaces/Aetna000/MemoryStackBench
Check it out here.
Please refer to the technical paper below.
https://github.com/aetna000/aetnamem/blob/e24d5aa74e62dbebeeb28a2e49b5c4d2006bbb4d/paper/aetnamem-control-plane.pdf
Check the benchmark
https://huggingface.co/spaces/Aetna000/MemoryStackBench
- 4 replies
replied to ZennyKenny's post 9 months ago
I have created https://huggingface.co/spaces/telcom/ResumeQA if someone wants to use or at least get some inspiration
reacted to ZennyKenny's post with 👍 9 months ago
Post
3,278
😎 My new personal website is live! Check out https://kennethhamilton.me to chat with an LLM about my professional skills and personal projects.
🙈 Think of it like a really, really vain version of ChatGPT.
🙈 Think of it like a really, really vain version of ChatGPT.
- 6 replies
replied to ZennyKenny's post 9 months ago
love the idea. I heard about this approach in https://www.youtube.com/watch?v=0teZqotpqT8
Stop Competing With 400 Applicants. Build This in One Weekend (Yes, there's a no code option too!)
reacted to AdinaY's post with 👍 9 months ago
Post
1,111
After a VLM, StepFun dropped a new audio model: Step-Audio-R1.1, enabling thinking while speaking 🔥
stepfun-ai/Step-Audio-R1.1
✨ Apache 2.0
✨ Combines dual-brain architecture and acoustic-grounded reasoning to enable real-time dialogue with SOTA-level reasoning
stepfun-ai/Step-Audio-R1.1
✨ Apache 2.0
✨ Combines dual-brain architecture and acoustic-grounded reasoning to enable real-time dialogue with SOTA-level reasoning
- 2 replies
posted an update 9 months ago
Post
1,591
MAD-GRPO: https://huggingface.co/blog/telcom/mad-grpo
In R1-Zero-Like Training *, Dr.GRPO treats GRPO’s by dropping std, but that often comes with a hidden side effect: length-weighted updates that can nudge model toward verbosity.
MAD-GRPO provides robust scale (MAD + epsilon) per-token normalization stability without verbosity bias.
*https://huggingface.co/papers/2503.20783
In R1-Zero-Like Training *, Dr.GRPO treats GRPO’s by dropping std, but that often comes with a hidden side effect: length-weighted updates that can nudge model toward verbosity.
MAD-GRPO provides robust scale (MAD + epsilon) per-token normalization stability without verbosity bias.
*https://huggingface.co/papers/2503.20783
reacted to their post with 👀 9 months ago
Post
215
CAQBI Index
is like a credit score for how an AI understands a concept.
Please check
https://huggingface.co/blog/telcom/caqbi
is like a credit score for how an AI understands a concept.
Please check
https://huggingface.co/blog/telcom/caqbi
- 1 reply
reacted to AdinaY's post with 👍 9 months ago
Post
1,550
Wechat AI is shipping!
WeDLM 🔥 A new language model that generates tokens in parallel, making it faster than standard LLMs , with the same Transformer setup!
https://huggingface.co/collections/tencent/wedlm
✨ 7B/8B - Base & Instruct
✨ Apache 2.0
WeDLM 🔥 A new language model that generates tokens in parallel, making it faster than standard LLMs , with the same Transformer setup!
https://huggingface.co/collections/tencent/wedlm
✨ 7B/8B - Base & Instruct
✨ Apache 2.0
- 4 replies
reacted to kanaria007's post with 👍 9 months ago
Post
2,010
✅ New Article: *PoC Architecture for Education & Developmental Support*
Title:
🎓 Building an SI-Core Wrapped Learning Companion - PoC architecture for education and developmental support
🔗 https://huggingface.co/blog/kanaria007/poc-architecture-for-education-development-support
---
Summary:
Most “AI tutors” are built as *LLM-first* systems. This article flips the default:
* The LLM is treated as an *untrusted proposal engine*
* *SI-Core owns* observation, consent, ethics, memory, and rollback
* Teachers and guardians get *real oversight*, not just chat transcripts
Scoped intentionally to *one subject × a small cohort (10–30 learners)*, this is a PoC you can actually ship—and audit.
> Don’t ask: “Can an AI replace teachers?”
> Prove: “Can we make an AI companion *safe, explainable, and governable* for real learners?”
---
Why It Matters (for AI on real stacks):
• *Consent & accommodations* are first-class (especially for minors / neurodivergent learners)
• *Ethics decisions are logged* (ALLOW / DENY / ESCALATE) with traceable reasoning
• “*Why this?*” explanations are built in for learners—and deeper inspection for adults
---
What’s Inside:
• A minimal reference architecture (frontend → SI-Gate → ethics/memory/logging → LLM APIs)
• Non-negotiables for the pilot (SI-wrapped LLM, Effect Ledger, ethics overlay, dashboards)
• Failure modes + safe-mode behavior
• Implementation checklist + rough effort/cost ballparks (kept explicitly non-normative)
---
📖 Structured Intelligence Engineering Series
A deployable pattern for taking today’s LLM tutor ideas and making them *auditable, overrideable, and rollback-safe*.
Title:
🎓 Building an SI-Core Wrapped Learning Companion - PoC architecture for education and developmental support
🔗 https://huggingface.co/blog/kanaria007/poc-architecture-for-education-development-support
---
Summary:
Most “AI tutors” are built as *LLM-first* systems. This article flips the default:
* The LLM is treated as an *untrusted proposal engine*
* *SI-Core owns* observation, consent, ethics, memory, and rollback
* Teachers and guardians get *real oversight*, not just chat transcripts
Scoped intentionally to *one subject × a small cohort (10–30 learners)*, this is a PoC you can actually ship—and audit.
> Don’t ask: “Can an AI replace teachers?”
> Prove: “Can we make an AI companion *safe, explainable, and governable* for real learners?”
---
Why It Matters (for AI on real stacks):
• *Consent & accommodations* are first-class (especially for minors / neurodivergent learners)
• *Ethics decisions are logged* (ALLOW / DENY / ESCALATE) with traceable reasoning
• “*Why this?*” explanations are built in for learners—and deeper inspection for adults
---
What’s Inside:
• A minimal reference architecture (frontend → SI-Gate → ethics/memory/logging → LLM APIs)
• Non-negotiables for the pilot (SI-wrapped LLM, Effect Ledger, ethics overlay, dashboards)
• Failure modes + safe-mode behavior
• Implementation checklist + rough effort/cost ballparks (kept explicitly non-normative)
---
📖 Structured Intelligence Engineering Series
A deployable pattern for taking today’s LLM tutor ideas and making them *auditable, overrideable, and rollback-safe*.
replied to their post 9 months ago