Amplitude analytics and Google Ads tracking detected without explicit privacy notice visible in provided content; Intellimize tracking enabled by default.
Terms of Service
—
No ToS or terms link observable in provided content.
Accessibility
+0.10
Article 27
MathJax support for mathematical content; responsive design evident; anti-flicker measures suggest consideration for user experience.
Mission
+0.05
Article 27
Together.ai focuses on AI accessibility and efficiency; no direct UDHR mission statement observable, but inference suggests pro-accessibility orientation.
Editorial Code
—
No editorial code of conduct observable.
Ownership
—
Together.ai is a private company; no observable UDHR-relevant ownership concerns.
Access Model
+0.08
Article 27
Blog content is freely accessible; no paywall; open access to technical knowledge supports Article 27 (participation in science and culture).
Ad/Tracking
-0.08
Article 12
Google Ads tracking, Amplitude analytics, and Intellimize conversion tracking present; multiple third-party tracking scripts load without explicit user consent mechanisms visible.
If this means there’s a 2x-7x speed up available to a scaled diffusion model like Inception Mercury, that’ll be a game changer. It feels 10x faster already…
Is anyone doing any form of diffusion language models that are actually practical to run today on the actual machine under my desk? There's loads of more "traditional" .gguf options (well, quants) that are practical even on shockingly weak hardware, and I've been seeing things that give me hope that diffusion is the next step forward, but so far it's all been early research prototypes.
I do wonder why diffusion models aren't used alongside constraint decoding for programming - surely it makes better sense then using an auto-regressive model.
Can't wait for the day I can actually try a diffusion model on my own machine (128GB M4 Max) rather than as a hosted service. So far I haven't seen a single piece of software that supports it.
I'd love to know what's going on with the Gemini Diffusion model - they had a preview last May and it was crazy fast but I've not heard anything since then.
A lot of this post-training recipe feels reminiscent of DINO training (teacher/student, use of stop gradients). I wonder if the more recent leJEPA SigREG regularization research might be relevant here for simpler post-training.
This doesn't mention the drawback of diffusion language models, the main reason why nobody is using them: they have significantly lower performance on benchmarks than autoregressive models at similar size.
Seeing half of an AR LLM's output tokens go to generating a predefined json schema bothers me so much. I would love to have an option to use diffusion for infilling.
Diffusion model papers are always interesting to read but I always feel like they need some mechanism to insert or delete tokens.
In the example in the figure in this post, once it has fixed "British munchkin cats _ _ and ..." you _can't_ get to "British munchkin cats are a new and controversial breed." because there's not the right number of tokens between "cats" and "and".
In a coding context, if your model samples a paren or a comma or something which is entirely plausible at that position, it can still close off an expansion which would be syntactically correct.
Score Breakdown
+0.15
PreamblePreamble
Low Practice
Editorial
ND
Structural
+0.15
SETL
ND
Combined
ND
Context Modifier
ND
Free public access to technical knowledge supports foundational UDHR principles of human dignity and universal rights, though no explicit rights advocacy present.
+0.10
Article 1Freedom, Equality, Brotherhood
Low Practice
Editorial
ND
Structural
+0.10
SETL
ND
Combined
ND
Context Modifier
ND
Public sharing of technical knowledge supports universal principles; no content suggests discrimination.
+0.10
Article 2Non-Discrimination
Low Practice
Editorial
ND
Structural
+0.10
SETL
ND
Combined
ND
Context Modifier
ND
No discriminatory practices observable; open access regardless of demographic factors.
ND
Article 3Life, Liberty, Security
No observable content relating to security or life.
ND
Article 4No Slavery
Not applicable to technical blog content.
ND
Article 5No Torture
Not applicable to technical blog content.
ND
Article 6Legal Personhood
Not applicable to technical blog content.
ND
Article 7Equality Before Law
Not applicable to technical blog content.
ND
Article 8Right to Remedy
Not applicable to technical blog content.
ND
Article 9No Arbitrary Detention
Not applicable to technical blog content.
ND
Article 10Fair Hearing
Not applicable to technical blog content.
ND
Article 11Presumption of Innocence
Not applicable to technical blog content.
-0.30
Article 12Privacy
Medium Practice Coverage
Editorial
-0.20
Structural
-0.25
SETL
+0.11
Combined
ND
Context Modifier
ND
Multiple tracking systems (Amplitude, Google Ads, Intellimize) actively collecting user data without observable explicit consent mechanism. Article 12 protects privacy; tracking contradicts this principle.
+0.20
Article 13Freedom of Movement
Medium Practice
Editorial
ND
Structural
+0.20
SETL
ND
Combined
ND
Context Modifier
ND
Free public access to technical content across borders supports freedom of movement and information.
ND
Article 14Asylum
Not applicable to technical blog content.
ND
Article 15Nationality
Not applicable to technical blog content.
ND
Article 16Marriage & Family
Not applicable to technical blog content.
ND
Article 17Property
Not applicable to technical blog content.
ND
Article 18Freedom of Thought
Not applicable to technical blog content.
+0.38
Article 19Freedom of Expression
Medium Advocacy Practice
Editorial
+0.45
Structural
+0.30
SETL
+0.26
Combined
ND
Context Modifier
ND
Technical blog post publishing research promotes freedom of opinion and expression; open dissemination of knowledge supports Article 19.
ND
Article 20Assembly & Association
Not applicable to technical blog content.
ND
Article 21Political Participation
Not applicable to technical blog content.
ND
Article 22Social Security
Not applicable to technical blog content.
ND
Article 23Work & Equal Pay
Not applicable to technical blog content.
ND
Article 24Rest & Leisure
Not applicable to technical blog content.
ND
Article 25Standard of Living
Not applicable to technical blog content.
ND
Article 26Education
Not applicable to technical blog content.
+0.52
Article 27Cultural Participation
Medium Advocacy Practice
Editorial
+0.50
Structural
+0.35
SETL
+0.27
Combined
ND
Context Modifier
ND
Blog post directly promotes scientific advancement and participation in culture/science; free access to technical knowledge and research results strongly supports Article 27. MathJax support enhances accessibility.