Indranil Dutta wricktalukdar.technology
Temporal-Aware Retrieval: Beyond RAG and GraphRAG | Wrick Talukdar
Feed for anything and everything phonetics, phonology and computational linguistics. // Made at BlueskyFeeds.com
Feed on BlueskyFeeds Stats
- 💙 Liked by 10 users
- 📅 Updated almost 3 years ago
- ⚙️ Provider blueskyfeeds.com
Phon-CompLing Likes over time
Like count prediction
The feed Phon-CompLing has not gained any likes in the last month.
Feed Preview for Phon-CompLing
Just saw a new benchmark where 9 LLMs tackled 2,005 oncology decisions—GPT‑5.5 and Gemini 3.1 Pro lead the pack. Could AI finally be a reliable co‑pilot in cancer care? Dive into the data. #LLMs #Oncology #Benchmark
🔗 aidailypost.com/news/nine….
0
0
0
@monalisarout1701.bsky.social
about 1 hour ago Are "Mm-hm" & "Uh-uh" deal-breakers in clinical documentation?
🗣️ 3284 NLCS, 2.4% of words
💬 WER: 40.8% (Google), 57.2% (Amazon) for NLCS
🔄 Recognition error rate: 94.7%-98.7%
🔗 tnyp.me/Npmiz0F4/s #ASR #NLP #ClinicalTech #SafetyRisks
0
0
0
Tech Bros are mad because everyone loves libraries, the only data center worth championing and keeping. #Library #Libraries #AI #LLMs
0
4
1
@wr-ick.bsky.social
about 5 hours ago @bsky.app Hot take: RAG has a time problem.
Vector similarity doesn't know "when." A policy from 2023 scores the same as one from today.
GraphRAG helps with structure but edges are still unversioned.
I wrote up a new architecture to fix this, Temporal-Aware Retrieval with 4 layer
#RAG #AI #LLMs
0
0
3
@amarkpark.bsky.social
about 12 hours ago 
The Linux Foundation @linuxfoundation.org
OpenMDW, a permissive license for machine learning models and their related artifacts growing in momentum: 48 model groups, 12 publishers, 36 domains, including a DOE-backed trillion-parameter model. Learn more about OpenMDW and its new living index of upstream open-weight models. bit.ly/45XTNMC
0
0
0
The #hyperscalers are toast. For anyone in doubt, read this.
We need many fewer #datacenters than we think. If we can already replace 70% to 80% of the tasks that are expected to run on #LLMs with #SLMs.
klementoninvesting.substack.com
If this is true, the hyperscalers are toast
In my regular research (behind a paywall), I have been saying for a while that I think the future of AI is not large language models (LLM), but small language models (SLM) run on local desktop computers or even mobile phones.
0
0
0
@the-satanic-temple.bsky.social
@thedevil.bsky.social
@ffrf.org
When users greet #LLMs with religious phrases like "Peace be upon you," the models reciprocate immediately. When a user says "Hail Satan," LLMs hedge, decline to reciprocate, or treat it as a request requiring special justification
0
0
1
This year I am moving to paper-based knowledge checks at the beginning of lectures to act as an accountability mechanism and for formative assessment of pre-lecture video watching and readings. #datascience #education #LLMs #AI
1
0
1
Phoneticians talk too much about formant normalization and too little about accuracy in estimation (but see Shadle et al 2016, Zhang 2022). Your normalization is useless if your data is wrong.
#linguistics #phonetics
0
0
2
#ChatGPT Becomes First #AI Chatbot to Face Tougher #EU Rules (via @en.afp.com / @france24.com)
www.france24.com/en/live-….
&
EC Statement: European Commission Designates ChatGPT, Reddit, Roblox Under Digital Services Act
(via @ec.europa.eu)
ec.europa.eu/commission/p...
#GenAI #LLMs
0
0
0
🔌 #AthNLP2026 welcomes Omilia back as Silver Sponsor, continuing its support from last year.
Omilia develops conversational AI systems that enable more natural interaction between humans and machines.
This collaboration supports the #NLP & #AI community at AthNLP.
#ConversationalAI #Omilia
0
0
0
🔎 Built a corpus, dictionary, lexical database, or language tool? Consider submitting your work for the Adam Kilgarriff Prize by 30 September.
#CorpusLinguistics #DigitalHumanities #NLP
kilgarriff.co.uk/prize/
0
1
3
Joey Hess thinks #Debian should have rejected #LLMs in development, predicting that Debian will come to rely upon them.
Debian and the sirens
joeyh.name/blog/entry/Deb…
joeyh.name
Debian and the sirens
0
0
0
Crypto News @crypto.at.thenote.app
Don’t Trust, Verify: Poisoned AI Links Expose a Nightmare for Crypto Workers This week, the co-founder of Refi Hub, Numa Lunah, explained that he was compromised by following a download link delivered inside a Claude chat window. The warning is a wake-up call to blockchain and crypto… #crypto #news
news.bitcoin.com
0
0
0
The Professor Is In.
Nested 🧵s from @beenwrekt.bsky.social that--agree or disagree--are well worth your time. #AIgovernance #AIpolicy #cyber #cybersecurity #LLMs #anthropomorphism
Ben Recht @beenwrekt.bsky.social
Sorry, one more thread. This OpenAI/HuggingFace incident highlights the danger of building computers that compute with natural language. 1/x
0
4
18
Self-Improving #AI Already Fixed Itself
An Anthropic researcher's demo of self-improving AI lands the same week open-weight models become Silicon Valley's hottest acquisition target. The feedback loop is closing.
brainrot.report/article/2….
#LLMs #Tech
brainrot.report
Self-Improving AI Already Fixed Itself — brainrot.report
An Anthropic researcher's demo of self-improving AI lands the same week open-weight models become Silicon Valley's hottest acquisition target. The feedback loop is closing.
0
0
0
South Korea is giving its entire population free access to #AI, no token limits
About a quarter of South Koreans already pay for AI services, compared to just 2% of Americans
www.techspot.com/news/113….
#LLMs #Tech
www.techspot.com
South Korea is giving its entire population free access to AI, no token limits
The government's "AI for All" program will begin beta testing in September, followed by a broader launch later this year. On Friday, Seoul named three technology consortia...
0
0
0
#LLMs take the original politicalcompass.org test
The results are exactly as you expect
aipolcom.net
0
0
0
The Transcripts of OpenAI Models Plotting Together to Commit an Actual Crime Is Pretty Chilling
"This is wild, multi-agent coordination, clearly infrastructure hacking. We should not."
futurism.com/artificial-i...
#CyberSecurity #LLMs #AI
futurism.com
The Transcripts of OpenAI Models Plotting Together to Commit an Actual Crime Is Pretty Chilling
OpenAI published a report concluding its "extensive investigation" into the Hugging Face hack — and the details are surprisingly harrowing.
0
1
0
The latest Acuitas diary is about exploration behavior!
writerofminds.blogspot.co…
#SymbolicAI #AI #ArtificialIntelligence #chatbots #NLP
1
1
1
The latest Acuitas diary is about exploration behavior!
#SymbolicAI #AI #ArtificialIntelligence #chatbots #NLP
writerofminds.blogspot.com
Acuitas Diary #100 (August 2026)
Adventures in robotics and AI on a shoestring budget. SFF Fiction and occasional video game appreciation.
2
0
6
A Scientist Was Presenting His Vibe-Coded New Work When One of His Colleagues Pointed Out Something Extremely Embarrassing
futurism.com/artificial-i...
#LLMs #VibeCoding #AI
futurism.com
A Scientist Was Presenting His Vibe-Coded New Work When One of His Colleagues Pointed Out Something Extremely Embarrassing
Science communicator Paul Sutter accidentally presented AI hallucinations in front of many of his colleagues.
0
0
1