Mati83moni/QUICK_JOBS
141 GB
Mateusz Piesiak PRO
Mati83moni
AI & ML interests
Hybrid LLMs, MoE, Mamba, multilingual NLP, biomedical reasoning, efficient pretraining
Recent Activity
published an article 5 minutes ago
Evaluation Awareness: One Direction, or Eight? updated a dataset 37 minutes ago
Mati83moni/evalaware-article-figures published a dataset 37 minutes ago
Mati83moni/evalaware-article-figuresOrganizations
None yet
Posts 1
Post
70
🔍 Watermark Finder — find, mark, and strip hidden text watermarks. Open source, MIT.
Detects six distinct covert channels with exact character offsets, not guesses: Unicode tag payloads, variation-selector encoding, zero-width binary/markers, homoglyph substitution, bidi controls, and repetition stamps. Also verifies C2PA content credentials (Adobe, Leica, Google-signed manifests) — and keeps "integrity" (did the bytes change?) strictly separate from "trust" (is the signer who they claim?), because a forged certificate verifies perfectly too.
It scores writing register (how much prose reads like unedited assistant output) with visible uncertainty ranges and per-feature contributions — and refuses a verdict below ~15 words rather than fake precision. That score measures register, not authorship, and the README says so louder than I'm saying it here.
Two deploy modes: a hosted engine (FastAPI on a HF Space) behind a Cloudflare Worker, or a fully client-side build compiled to WebAssembly via Pyodide — the document never leaves the browser tab in that mode.
304 tests across four suites, real CI on every push, no accounts (anonymous HMAC-token workspace).
Live:
https://watermark-finder.pages.dev/
Source:
https://github.com/Mati83mon/Watermark_Finder
Detects six distinct covert channels with exact character offsets, not guesses: Unicode tag payloads, variation-selector encoding, zero-width binary/markers, homoglyph substitution, bidi controls, and repetition stamps. Also verifies C2PA content credentials (Adobe, Leica, Google-signed manifests) — and keeps "integrity" (did the bytes change?) strictly separate from "trust" (is the signer who they claim?), because a forged certificate verifies perfectly too.
It scores writing register (how much prose reads like unedited assistant output) with visible uncertainty ranges and per-feature contributions — and refuses a verdict below ~15 words rather than fake precision. That score measures register, not authorship, and the README says so louder than I'm saying it here.
Two deploy modes: a hosted engine (FastAPI on a HF Space) behind a Cloudflare Worker, or a fully client-side build compiled to WebAssembly via Pyodide — the document never leaves the browser tab in that mode.
304 tests across four suites, real CI on every push, no accounts (anonymous HMAC-token workspace).
Live:
https://watermark-finder.pages.dev/
Source:
https://github.com/Mati83mon/Watermark_Finder
Articles 1
Article
Evaluation Awareness: One Direction, or Eight?
spaces 10
Running
Is Evaluation Awareness One Direction, or Eight?
🔬
Factor-controlled probe of evaluation-awareness geometry
Running
Evaluation Awareness in PLLuM-12B
🔬
One direction or eight? The same probe, run on PLLuM-12B
Paused
Text Provenance Lab
🔍
Watermark and text-provenance analysis engine
Running
Watermark Finder
🔍
Find, place and strip hidden text watermarks
Paused
Agents
Model Playground
⚡
Universal HF playground for text/image/audio/video
Paused
Agents
Custom Model Quantization
🗜
Submit a custom model quantization request
models 5
Mati83moni/PLLuM-12B-chat-2512-EXL3
Text Generation • 4B • Updated • 365
Mati83moni/HybridMoE-Titan-v2
Text Generation • Updated
Mati83moni/HybridMoE-Titan-v1
Text Generation • Updated
Mati83moni/Llama-PLLuM-70B-instruct-2512-GGUF
Text Generation • 71B • Updated • 52
Mati83moni/functiongemma-270m-it-mobile-actions
Text Generation • 0.3B • Updated • 19 • 1