volodia

5 ms·

1.
How to build a diffusion language model(kuleshov-group.github.io)
184 points by volodia 18d ago|20 comments
2.
Next-Edit in Kilo, Powered by Inception Diffusion LLMs(blog.kilo.ai)
2 points by volodia 3mo ago|0 comments
3.
Mercury 2 on PinchBench: Diffusion LLM benchmarked on real OpenClaw agent tasks(inceptionlabs.ai)
2 points by volodia 6mo ago|0 comments
4.
by volodia 7mo ago
Thank you for the detailed feedback! I shared this already with the team.
5.
by volodia 7mo ago
This looks like an inference glitch that we are working on fixing, thank you for flagging.
6.
by volodia 7mo ago
There are many ways to do it, but the simplest approach is block diffusion: https://m-arriola.com/bd3lms/ There are also more advanced approaches, for example FlexMDM, which essentially predicts length of the "can
7.
by volodia 7mo ago
Would love to hear about your experience. Send us an email.
8.
by volodia 7mo ago
Not imminently, but hard to predict where the field will go
9.
by volodia 7mo ago
There are few: fast agents, deep research, real-time voice, coding. The other thing is that when you have a fast reasoning model, you spend more effort on thinking in the same latency budget, which pushed up quality.
10.
by volodia 7mo ago
We agree! In fact, there is an emerging class of models aimed at fast agentic iteration (think of Composer, the Flash versions of proprietary and open models). We position Mercury 2 as a strong model in this category.
11.
by volodia 7mo ago
That is also our view! We see Mercury 2 as enabling very fast iteration for agentic tasks. A single shot at a problem might be less accurate, but because the model has a shorter execution time, it enables users to iterate much more quickly.
12.
by volodia 7mo ago
You can think of Mercury 2 as roughly in the same intelligence tier as other speed-optimized models (e.g., Haiku 4.5, Grok Fast, GPT-Mini–class systems). The main differentiator is latency — it’s ~5× faster at comparable quality. We’re not
13.
by volodia 7mo ago
Thanks for trying it and for the thoughtful feedback, really appreciate it. And we’re actively working on improving quality further as we scale the models.
14.
by volodia 7mo ago
Thank you for your patience. We are working to handle the surge in demand.
15.
by volodia 7mo ago
Just to clarify one point: Mercury (the original v1, non-reasoning model) is already used in production in mainstream IDEs like Zed: https://zed.dev/blog/edit-prediction-providers Mercury v1 focused on autocomplete and
16.
by volodia 7mo ago
I’d push back a bit on the Pareto point. On speed/quality, diffusion has actually moved the frontier. At comparable quality levels, Mercury is >5× faster than similar AR models (including the ones referenced on the AA page). So for
17.
by volodia 7mo ago
Co-founder / Chief Scientist at Inception here. If helpful, I’m happy to answer technical questions about Mercury 2 or diffusion LMs more broadly.
18.
Mercury 2: Best-in-class speed-optimized intelligence at 1,200 tok/SEC(twitter.com)
1 points by volodia 7mo ago|0 comments
19.
by volodia 10mo ago
There is also this one that was released in October: https://github.com/kuleshov/char-mdlm
20.
by volodia 2y ago
the LLaDA paper is a scaled-up version of this paper; they cite it as an anonymous ICLR submission
21.
by volodia 2y ago
Great question! The model can more efficiently leverage existing GPU hardware---it performs more computation per unit of memory transferred; this means that on older hardware one should be able to get similar inference speeds as one would g
22.
by volodia 2y ago
Yes, we plan to be releasing a tech report soon. We are not open sourcing the models at launch time, but we have a roadmap of future releases in which we hope to make some of our models accessible to the research community.
23.
by volodia 2y ago
That's a good point. In this context, we've been using "commodity GPUs" to refer to standard Nvidia hardware, in contrast to specialized chips like Groq and Cerebras. While these chips also achieve fast speeds, they are
24.
by volodia 2y ago
Good question! We are not open sourcing the models at launch time, but we have a roadmap of future releases in which we hope to make some of our models accessible to the research community.
25.
by volodia 2y ago
The short answer is that we do more than one parallel pass over multiple tokens: we iteratively refine them over a few passes to fix incoherences. This can be seen as a generalization of diffusion algorithms that underlie systems like Midjo
26.
by volodia 2y ago
Not today, but we will be following up with a technical report over the next week or so. In the meantime, you can take a look at some of the research papers that inspired our work: - https://arxiv.org/abs/2310.16834 -
27.
by volodia 2y ago
This is Volodymyr, co-founder at Inception---let us know if you have any questions about diffusion, language modeling, and our new Mercury models!
28.
Finetuning 3-Bit LLMs on Consumer GPUs by Integrating with Modular Quantizers(arxiv.org)
2 points by volodia 3y ago|0 comments
29.
LLMTune: 4-Bit finetuning of 65B LLAMA models on a single consumer GPU(github.com)
3 points by volodia 3y ago|0 comments
30.
LLMTune: 4-Bit Finetuning of LLMs on a Consumer GPI(twitter.com)
2 points by volodia 3y ago|0 comments

More ›