#LLM-based
LightNews LightNews
arxiv-cs-cl.bsky.social
Song Jin, Shuqi Li, Shukun Zhang, Rui Yan
FinRpt: Dataset, Evaluation System and LLM-based Multi-agent Framework for Equity Research Report Generation
https://arxiv.org/abs/2511.07322
November 11, 2025 at 5:34 AM
researchtrend.ai
[2025-11-11] 📚 Updates in #ELM

(1) LPFQA: A Long-Tail Professional Forum-based Benchmark for LLM Evaluation
(2) Efficient LLM Safety Evaluation through Multi-Agent Debate

🔍 More at researchtrend.ai/communities/ELM
November 11, 2025 at 4:02 AM
researchtrend.ai
[2025-11-11] 📚 Updates in #ALM

(1) <a href="https://researchtrend.ai/papers/2511.06346" class="hover:underline text-blue-600 dark:text-sky-400 no-card-link" target="_blank" rel="noopener" data-link="bsky">LPFQA: A Long-Tail Professional Forum-based Benchmark for LLM Evaluation
(2) LPFQA: A Long-Tail Professional Forum-based Benchmark for LLM Evaluation

🔍 More at researchtrend.ai/communities/ALM
November 11, 2025 at 4:02 AM
aixbiobot.bsky.social
Can Lightweight LLM Agents Improve Spatial TranscriptomicsAnnotation? [new]
Lightweight LLMs annotated spatial transcriptomics data by integrating rule-based heuristics and multi-role reasoning within an agentic framework.
Can Lightweight LLM Agents Improve Spatial TranscriptomicsAnnotation? Table 1 Table 2 Table 8
November 10, 2025 at 10:54 PM
ryytikki.bsky.social
yes, and this has basically 0 of the negatives of LLM based things (insane power/water usage, stolen training data, etc)

This technique has been used for adaptive animations for years. It lets you have movement that fits the world better without an insane and unfeasable amount of tedious work
November 10, 2025 at 9:21 PM
matasoft.bsky.social
Real spreadsheet superpowers: summarize information, extract trends, and verify external sources automatically—thanks to connected SearXNG search and powerful cloud-based LLMs. #AI #LLM #Ollama #OllamaCloud #Searxng #UnPerplexedSpready
matasoft.hr/qtrendcontro...
(Un)Perplexed Spready with web search enabled Ollama models
Configure web search-enabled Ollama models in (Un)Perplexed Spready. Complete tutorial for connecting SearXNG to local AI models for enhanced spreadsheet automation.
matasoft.hr
November 10, 2025 at 8:26 PM
wisemom.bsky.social
Oh gosh…
Most (all?) attorney rules of professional conduct require due diligence…so at least checking that citations are real!!
Plus, I would be concerned about feeding confidential data into a LLM/IA which is based on data mining.

These are just a few problems with using AI in the legal field.
November 10, 2025 at 7:34 PM
nico-encounter.bsky.social
Some more thoughts about this. Even with continuous training, an LLM based AI would always be equally good at mimicking the text of its conversation partner (as measured by training loss) as its own past outputs, before post training.
November 10, 2025 at 6:50 PM
willtor.bsky.social
It looks like they're reasoning because the information they were trained on was written by people who reasoned. But they're regurgitating what they've consumed.

Maybe one day AI will reason, but if it does, it won't be based on LLM technology.
November 10, 2025 at 6:30 PM
firetailai.bsky.social
WhisperLeak is vulnerability that affects streaming languages based on a fundamental characteristic rather than an implementation flaw.

Read here: thehackernews.com/2025/11/micr...

Keep up with AI breaches here: www.firetail.ai/ai-breach-tr...

#WhisperLeak #AI #AISecurity #LLM #Breaches #FireTail
November 10, 2025 at 5:22 PM
ingorohlfing.bsky.social
to me what the added value is in 2025. It discusses standard prompting w/o any insights into how students liked LLM-based training or so.
Two other points:
1) If one uses an LLM, I am not sure why one would use Julius. The free account does not suffice for teaching bc of a prompt limit. 2/
November 10, 2025 at 4:47 PM
jpiaskowski.bsky.social
These graphs of self-taught teenagers' startup to design agricultural herbicides is something else. 'Bindwell’s website once described these [AI] models as insanely fast [that] could predict, in mere seconds, the results of experiments that would have taken days' www.wired.com/story/ai-sta...
Anand grew up in India reading Hacker News with his dad and was building his own large language models halfway through high school. Before he graduated, he, his cofounder (now 18), and two other friends from summer camp published a paper on bioRxiv, about an LLM they’d built to predict a facet of protein behavior. It got scientists buzzing on X. The paper was cited in a well-respected journal. They decided they should try to build a startup, brainstormed, and settled on protein-based pesticides. Then, the fairy tale continues, a wood sprite (sorry, venture capitalist) got in touch on LinkedIn and offered them $750,000 to drop out of high school and college and work on the company full-time. They accepted and got started. The teens knew next to nothing about agribusiness. That was last December. Five months later, Anand and his cofounder opened their first biological testing lab in the San Francisco Bay Area, then moved to another, where they personally squeeze drops of promising molecules into tiny vials. (A protein-based compound can more precisely target a locust or aphid, goes the theory, and not also take out the humans, earthworms, bees.) I asked him how he’d picked up the skills to work in a wet lab. “I hired a friend,” he said cheerfully. The friend coached him over the summer before heading back to college in the fall. “Now I can do some biochemical assays,” Anand says. “Not like a whole range of assays, but basic, wet-lab validation of our models.” Huh, I thought. That a few teens had in a handful of months built their own LLMs, learned the biochemistry of pest control, used their models to identify potential molecules, and were now pipetting away in their own lab seemed not shabby. In truth, once I’d tallied up all that they’d done, it struck me as completely absurd.
November 10, 2025 at 4:06 PM
iresiragusa.bsky.social
📄 An automatic alignment strategy for medical ontologies: the MedPix 2.0 case study

Our latest work presented at HC@AIxIA + HYDRA 2025 workshop in #ECAI2025

We proposed an ontological version of ICD-11 and MedPix 2.0 data set and an LLM-based pipeline for automatic alignment
November 10, 2025 at 9:46 AM
shmmeee.bsky.social
Anyone offering an LLM based triage service? If not I might have an idea…
November 10, 2025 at 9:21 AM
arxiv-cs-cl.bsky.social
Yu Li, Lehui Li, Qingmin Liao, Fengli Xu, Yong Li
AgentExpt: Automating AI Experiment Design with LLM-based Resource Retrieval Agent
https://arxiv.org/abs/2511.04921
November 10, 2025 at 7:38 AM
ecoevorxiv.bsky.social
"Emerging Applications of Large Language Models in Ecology and Conservation Science"
doi.org/10.32942/X2P...
Countries contributing to research on LLM applications in ecology and conservation science, based on the distribution of author affiliations in the reviewed articles, preprints, and conference proceedings. The USA, China, the UK, Australia, and Germany were the top five contributors.
November 10, 2025 at 7:30 AM
cscl-bot.bsky.social
Yu Li, Lehui Li, Qingmin Liao, Fengli Xu, Yong Li: AgentExpt: Automating AI Experiment Design with LLM-based Resource Retrieval Agent https://arxiv.org/abs/2511.04921 https://arxiv.org/pdf/2511.04921 https://arxiv.org/html/2511.04921
November 10, 2025 at 6:30 AM
emnlpmeeting.bsky.social
Senior Area Chair Highlights (4/9)
ReSo: A Reward-driven Self-organizing LLM-based Multi-Agent System for Reasoning Tasks Heng Zhou, Hejia Geng, Xiangyuan Xue, Li Kang, Yiran Qin, Zhiyong Wang, Zhenfei Yin, LEI BAI Prototypical Human-AI Collaboration Behaviors from LLM-Assisted Writing in the Wild Sheshera Mysore, Debarati Das, Hancheng Cao, Bahareh Sarrafzadeh Comparing human and LLM politeness strategies in free production Haoran Zhao, Robert D. Hawkins Who Holds the Pen? Caricature and Perspective in LLM Retellings of History Lubna Zahan Lamia, Mabsur Fatin Bin Hossain, Md Mosaddek Khan
November 10, 2025 at 3:47 AM
azurelionprod.bsky.social
Indeed! AI bro prompters are, currently, useful idiots for the tech bro oligarchs.
Their usefulness will end.
Isn't AI, in at least some cases, already creating its own prompts?
Wetterschneider @Stretchedwiener · 2h AI generative software is just a mechanism to put the generation of mimicked output directly in the hands of the "money people". No auteurs needed, no crew, no craftspeople, no artists. No-one. Just a response to viewer metrics fed back into a system to generate more empty content. All the AIbros bleating about "democratization" and "gatekeeping" don't seem to realize that they aren't needed either, in the end game. They are just beta-testers, ham-fistedly churning out test cases. They think actual artists aren't valuable but they refuse to see that they have even less to offer. Netflix is engineering a feedback system where 5%-10%-20% of its content is generated automatically based on viewer metrics. So is YouTube. They all are. No creatives in the pipe. And no "prompters" either. Prompters are even easier to replace with an LLM. And people will watch. Viewers (some very young) are already being exposed to a chain of ever degrading content automatically queued on YouTube.
November 9, 2025 at 5:49 PM
bb.wavelengths.online
thinking about starting a company based around hyper-specific LLM-based cash flow optimization strategies for small-to-medium-sized real estate brokerage advertising firms looking to expand their reach into home security sourcing infrastructure.

hmu if that sounds chill
November 9, 2025 at 5:21 PM
sesheta.bsky.social
So all this to say that while the current “AI” push obviously appeals to right-wingers wanting to create an alternative-fact-based propaganda LLM, that they by definition are doomed make a moronic mechahitler, while models accounting for reality can potentially contribute to cancer research, etc. 6/
November 9, 2025 at 4:59 PM
blastingpowder.bsky.social
Outlandish claims of LLM productivity boosts are based on gains in quality metrics that measure only activities they can reasonably do. It's like trying to prove in 1920 that e.g airplanes will reshape all human economic activity by pointing at consistent gains in aerodynamics or fuel efficiency
November 9, 2025 at 4:32 PM
bvimarlins.bsky.social
For coding the local ML/LLM products have kept pace with the API based ones this entire time, which has only added to the ridiculousness of this hype cycle imo
November 9, 2025 at 1:16 PM
flavianvespasian.bsky.social
Fancy chatbot scam coming to an end?

The sci-fi based delusions of Silicon Valley morons hitting a wall of reality?

A gagillion data centers will never make LLM sentient?

Next, Tesla shareholders find out that iRobot Optimus is a fantasy apartheid boy saw in a Will Smith movie?
November 9, 2025 at 1:00 PM
lorak.bsky.social
in what way do you use & think llm based products make you more productive
November 9, 2025 at 12:53 PM