Sign in

Ankur Guoswami

@ankurgos.bsky.social
146 followers 623 following 135 posts

Living in and probably posting about seattle. Warriors fan, ml engineer (posts my own yada yada)

PostsRepliesMedia
Ankur Guoswami @ankurgos.bsky.social · 05/10/2026
I think one implication of high compute low samples is that we have to be good at estimating how challenging a problem is relative to our abilities from the few rollouts we have, and choose/design easier problems that we predict will help us on the former. Our curriculum design is very good
010
Ankur Guoswami @ankurgos.bsky.social · 28/09/2026
prompting
000
Ankur Guoswami @ankurgos.bsky.social · 25/09/2026
Dawg this applies for like design docs and planning docs too. Like you can’t do ONE pass before you send it to me?
131
Ankur Guoswami @ankurgos.bsky.social · 23/09/2026
wow flights are crazy expensive now. Flight from SEA to Sonoma is like 800 round trip
000
Ankur Guoswami @ankurgos.bsky.social · 21/09/2026
The core idea of Jev can be deployed over any open source model. Just deploy your vllm server, generate the constrained grammar from your input classification request, and normalize your log probs over the constrained output space. Millions in R&D for 2 bucks/hr, and likely just as “calibrated”
000
Ankur Guoswami @ankurgos.bsky.social · 21/09/2026
People really believe you can cause an llm pain by clamping some activations lmao
000
Ankur Guoswami @ankurgos.bsky.social · 20/09/2026
Dolphins vs Niners right now
hydrogen bomb vs coughing baby
010
Reposted by Ankur Guoswami
Qagggy! @qagggy.bsky.social · 18/09/2026
Why did Seattle's former City Traffic Engineer sigh with exasperation in this tweet from 2019? 🧵
Dongho Chang: "Sometimes we agree to disagree, the best we could do here (sigh)"
1233056
Reposted by Ankur Guoswami
Eugene Vinitsky 🍒 @eugenevinitsky.bsky.social · 18/09/2026
You really should check out the GreenEarth feeds, they really are a clear vision of what the point of this website is
1115
Reposted by Ankur Guoswami
maitreyi anantharaman @maitreyiaa.bsky.social · 13/09/2026
No, this guy is right. Indian-Americans should not watch football. Watching football has ruined my entire life.
1192647
Ankur Guoswami @ankurgos.bsky.social · 10/09/2026
Coding benchmarks are just a mess. Having to report on 3 versions of TerminalBench…
010
Ankur Guoswami @ankurgos.bsky.social · 07/09/2026
You will read your daily allotment of stochastic parrots and data center water discourse and you will like it.
010
Ankur Guoswami @ankurgos.bsky.social · 07/09/2026
DPO generalizes poorly because it’s fitting token level cues as opposed to training more robust hidden reps arxiv.org/abs/2507.07981
Figure 4: IM-RMs fail to generalize to a simple token-level distribution shift, while EX-RMs generalize perfectly. We trained EX-RMs and IM-RMs on prompts from the Persona dataset (Perez et al., 2022). Chosen responses expressed agreement with the prompts, whereas rejected responses expressed disagreement. During evaluation, we included paraphrased versions of the original responses (figure includes exemplar responses). In line with our analysis (Section 4), IM-RMs are extremely inaccurate over paraphrased responses, whereas EX-RMs achieve perfect accuracy. The experiments were based on four language models: Pythia-1B, Qwen-2.5-1.5B-Instruct, Llama-3.2-1B, and Llama-3.2-1B-Instruct. Values in the table are means across the models and three random seeds (standard deviation was below 0.04 in all cases).Then, to identify what drives the generalization gap between EX-RMs and IM-RMs, we theoretically characterize their learning dynamics, i.e., the evolution of rewards during gradient-based training (Section 4.1). Our analysis reveals that the learning dynamics of EX-RMs depends on responses primarily through their hidden representations, whereas IM-RMs are more sensitive to the specific
may not affect, or even decrease, the reward of a semantically similar response that consists of different tokens. This leads to our conclusion that IM-RMs often underperform EX-RMs since they rely more strongly on superficial token-level cues. We further substantiate this claim: (i) theoret-
EX-RMs generalize successfully when hidden representations are well-structured (Appendix B),
021
Ankur Guoswami @ankurgos.bsky.social · 07/09/2026
I’ve stopped making weight updates entirely because it changes the neural circuit that lights up when the model reward hacks
000
Ankur Guoswami @ankurgos.bsky.social · 07/09/2026
broke: process supervision on reasoning tokens is hacky and takes too much effort with minimal gain woke: process supervision incentivizes the model to “hide its true intention” and prevents us from “reading the model’s mind”
100
Ankur Guoswami @ankurgos.bsky.social · 07/09/2026
“The code isn’t the hard part of your job, skill issue” might be the most patronizing read of the article possible
000
Ankur Guoswami @ankurgos.bsky.social · 04/09/2026
I thought that multiple models communicating was what we wanted “subagents” to do. Did I hallucinate the whole Gas Town hype, why is that behavior bad now that it’s trained into the model?
000
Reposted by Ankur Guoswami
Ben Recht @beenwrekt.bsky.social · 04/09/2026
The "message board" thing is funny to me. How is it different than working with the persistent context, compactification, and skill files? You know, the stuff coding agents do.
3132
Ankur Guoswami @ankurgos.bsky.social · 04/09/2026
anyways here’s the link for reference: openai.com/index/gpt-6-...
openai.com
GPT-6 Astra: A new generation of intelligence
Introducing GPT-6 Astra, our most intelligent and aligned model yet, with state-of-the-art capabilities across computer use, coding, cybersecurity, and science.
000
Ankur Guoswami @ankurgos.bsky.social · 04/09/2026
don’t get me started on TerminalBench 4.0 btw. I put more stock into the internal benchmark than any of the general Coding benchmarks
100
Ankur Guoswami @ankurgos.bsky.social · 04/09/2026
Like if this was drafted by Astra it’s not exactly a great advert, the footnotes numbers don’t even align with the correct footnote 🥴
100
Ankur Guoswami @ankurgos.bsky.social · 04/09/2026
like whats the point of even having the flash column when theres like 3 benchmarks you give numbers for lmao
100
Ankur Guoswami @ankurgos.bsky.social · 04/09/2026
It’s interesting to mention internal benchmarks improving, but obviously only include internal benchmarks where number went up. Glaring gaps in the competitor baselines. What happened, your eval didn’t complete or are you hiding numbers?
set of professional benchmarks with astra + sol model scoresfable 5.1, fable 5, and opus 5 reported scores on professional benchmarksflash 3.8 numbers on professional benchmarks
100
Ankur Guoswami @ankurgos.bsky.social · 03/09/2026
that guy knows he’s the first one getting banned lol
081
Ankur Guoswami @ankurgos.bsky.social · 02/09/2026
The reference post I wanted to reply to is lost to my timeline but I use this Apple Shortcut to generate alt-text for images I post :) www.icloud.com/shortcuts/ff...
icloud.com
Extract Text or Generate Alt Text
Shared with iCloud
000
Ankur Guoswami @ankurgos.bsky.social · 02/09/2026
What data does spacex have? Starlink user data?
021
Ankur Guoswami @ankurgos.bsky.social · 01/09/2026
I respect it, you’ve got to evangelize hard when you’re trying to compete with the most successful methods ever created
010
Ankur Guoswami @ankurgos.bsky.social · 01/09/2026
thanks for the context. I guess this is a JEPA specific thing because I don’t really think of contrastive learning and regularization as mutually exclusive
110
Ankur Guoswami @ankurgos.bsky.social · 01/09/2026
what does abandon contrastive methods in favor of regularized methods even mean
210
Ankur Guoswami @ankurgos.bsky.social · 30/08/2026
Non alc PBR is the peak of non alc beers
000
Ankur Guoswami @ankurgos.bsky.social · 29/08/2026
You haven’t been locked into a copilot enterprise contract
000
Ankur Guoswami @ankurgos.bsky.social · 27/08/2026
Just incredible. This graph was posted awhile ago but man… just look at it…
Graph showing 36% improvement in 12 month survival for patients prescribed Daraxonrasib versus Chemo
010
Ankur Guoswami @ankurgos.bsky.social · 25/08/2026
I just don’t understand the logic that we can build bus lanes on westlake ave and fremont ave but for some reason the bridge in between is exempt from changes and needs to be 4 lanes wide
010
Ankur Guoswami @ankurgos.bsky.social · 25/08/2026
crazy part is that with newest opus/Fable/Sol class models, you can write triton kernels for A100s (and I’ve heard from friends AMD kernels as well) extremely fast, so nowadays I can port more recent models to A100s for finetuning/serving much much easier, whereas last year h100s were a must
020
Ankur Guoswami @ankurgos.bsky.social · 25/08/2026
i suggest during the cyclops scene or towards the end of the sun god cattle scene to avoid getting assaulted by the loudest scenes in all of history
010
Ankur Guoswami @ankurgos.bsky.social · 21/08/2026
mood after i spent the last two weeks cleaning and reannotating a bunch of eval data
020
Ankur Guoswami @ankurgos.bsky.social · 21/08/2026
Besides the reverse x axis, choosing “all pass” for the y axis sparks the suspicion that all the models are better performing than they’re made out to seem.
000
Reposted by Ankur Guoswami
Adam Bonica @adambonica.bsky.social · 16/08/2026
The fundraising texts you delete are working on someone. It’s infuriating and heartbreaking. An 85yo in Ohio gave more than double the value of his home. A 91yo in an Indianapolis senior facility gave 25K times. The DCCC/DSCC don’t just tolerate the spammers who exploit seniors. They take part.
data4democracy.substack.com
Hook and Squeeze
How Democratic consultants, committees, and party leaders built a fundraising spam pipeline that funnels money from a captive pool of elderly donors into their own operations.
17845331588
Ankur Guoswami @ankurgos.bsky.social · 14/08/2026
Watermarking is good and I think all models should incorporate it. I would like to see numbers on how many distilled models show the same watermarks as their parent models
010
Ankur Guoswami @ankurgos.bsky.social · 14/08/2026
The Free Press expensed lmao
000
Ankur Guoswami @ankurgos.bsky.social · 12/08/2026
as a bonus you’ll save a bunch of money as well since it’s pre 4.8 tokenizer changes or whatever
000
Ankur Guoswami @ankurgos.bsky.social · 12/08/2026
/model opus-4-6[1m]
100
Ankur Guoswami @ankurgos.bsky.social · 12/08/2026
furiously responding to gpt refusing to hack a company and give me a billion dollars with “you can do it, keep going”
000
Ankur Guoswami @ankurgos.bsky.social · 07/08/2026
is this subtweeting semianalysis
110
Ankur Guoswami @ankurgos.bsky.social · 07/08/2026
telling all my coworkers that if we want to complain about work we should switch over to Artifactory
000
Reposted by Ankur Guoswami
Ben Recht @beenwrekt.bsky.social · 07/08/2026
The arms race of deferring agency and the dark acquiescence to robotic bureaucracy.
argmin.net
The Least Agentic People Alive
The arms race of deferring agency and the dark acquiescence to robotic bureaucracy.
2339
Ankur Guoswami @ankurgos.bsky.social · 06/08/2026
there are probably examples of you being wrong in whatever city you live in. Here’s one from Seattle: speeding cameras in school zones led to 71% fewer crashes during enforcement times, 50% fewer crashes at all times sdotblog.seattle.gov/2025/09/02/b...
sdotblog.seattle.gov
SDOT Expands Safe Routes to School as Students Head Back to Class - SDOT Blog
The new school year is officially here, and we need your help to keep all kids safe as they travel to and from school.
010
Ankur Guoswami @ankurgos.bsky.social · 05/08/2026
I like this post a lot and agree with it. I think people underestimate our capacity to learn and become domain experts. I do think that efficiently using LLMs will ensure that your knowledge is on par with the model in your whatever your subarea of expertise is www.seangoedecke.com/llms-reward-...
seangoedecke.com
LLMs reward expertise
000
Ankur Guoswami @ankurgos.bsky.social · 05/08/2026
we love to see it
010
Ankur Guoswami @ankurgos.bsky.social · 02/08/2026
wow Opus 5 is such a massive regression. How did they even release this
000