AI & ML interests
OpenFree_AI
Recent Activity
View all activity
javadtaghia
posted an update about 10 hours ago
javadtaghia
posted an update 8 days ago
Post
1,099
Hugging face article: https://huggingface.co/blog/javadtaghia/atmem-2310-bounded-evidence-discovery-for-persiste
AtMem 2.3.10 is out; and our full DolphinBench result has more than doubled.
AtMem 2.3.9 passed 192 of 600 tasks (32%).
AtMem 2.3.10 passed 390 of 600 (65%). We doubled it.
What changed?
We focused on how an agent discovers and uses evidence.
AtMem combined lexical and semantic retrieval while keeping related passages together with their original source and provenance and more.
Where do we stand?
This is substantial progress, with a clear tradeoff: median recorded task latency increased from 31.87 to 85.78 seconds. More successful investigation currently takes more time.
The published Hermes/Luna/Mem0 reference scores 424/600 (70.7%).
What comes next?
We want to understand the remaining failures: was evidence missing, poorly ranked, packed incompletely, interpreted with the wrong temporal state, or used incorrectly during an action?
Evidence discovery remains opt-in.
Release: https://github.com/aetna000/atmem/releases/tag/v2.3.10
DolphinBench submission PR: https://github.com/mem0ai/dolphinbench/pull/10
Results and execution evidence: atmem/atmem-2.3.10b1-dolphinbench
PS: DolphinBench tests whether an agent can use a simulated user’s history to complete actions in applications. Finding a relevant passage isn’t enough—a task passes only when all its required checks pass.
AtMem 2.3.10 is out; and our full DolphinBench result has more than doubled.
AtMem 2.3.9 passed 192 of 600 tasks (32%).
AtMem 2.3.10 passed 390 of 600 (65%). We doubled it.
What changed?
We focused on how an agent discovers and uses evidence.
AtMem combined lexical and semantic retrieval while keeping related passages together with their original source and provenance and more.
Where do we stand?
This is substantial progress, with a clear tradeoff: median recorded task latency increased from 31.87 to 85.78 seconds. More successful investigation currently takes more time.
The published Hermes/Luna/Mem0 reference scores 424/600 (70.7%).
What comes next?
We want to understand the remaining failures: was evidence missing, poorly ranked, packed incompletely, interpreted with the wrong temporal state, or used incorrectly during an action?
Evidence discovery remains opt-in.
Release: https://github.com/aetna000/atmem/releases/tag/v2.3.10
DolphinBench submission PR: https://github.com/mem0ai/dolphinbench/pull/10
Results and execution evidence: atmem/atmem-2.3.10b1-dolphinbench
PS: DolphinBench tests whether an agent can use a simulated user’s history to complete actions in applications. Finding a relevant passage isn’t enough—a task passes only when all its required checks pass.
- 12 replies
javadtaghia
posted an update 11 days ago
Post
107
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
javadtaghia
posted an update 15 days ago
Post
141
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
javadtaghia
posted an update 22 days ago
Post
73
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
javadtaghia
posted an update 23 days ago
Post
116
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
javadtaghia
posted an update about 1 month ago
Post
129
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.
javadtaghia
posted an update about 1 month ago
Post
120
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?
iproskurina
authored 2
papers about 2 months ago
iproskurina
submitted a
paper to Daily Papers about 2 months ago
seawolf2357
submitted a
paper to Daily Papers 2 months ago
iproskurina
authored a
paper 3 months ago
iproskurina
submitted a
paper to Daily Papers 3 months ago
iproskurina
authored 3
papers 6 months ago
Fair-GPTQ: Bias-Aware Quantization for Large Language Models
Paper • 2509.15206 • Published • 1
HatePrototypes: Interpretable and Transferable Representations for Implicit and Explicit Hate Speech Detection
Paper • 2511.06391 • Published • 1
RILEC: Detection and Generation of L1 Russian Interference Errors in English Learner Texts
Paper • 2603.07366 • Published
seawolf2357
submitted a
paper to Daily Papers 6 months ago
javadtaghia
posted an update 10 months ago
Post
232
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
javadtaghia
posted an update 10 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