Machine Learning @cohere | Previously PhD @InfAtEd, @MetaAI (FAIR) and @AmazonScience.
London, UK
Joined November 2016
- Tweets783
- Following504
- Followers524
- Likes4.6K
Christos Baziotis retweeted
Fresh list of 300+ startup jobs in Greece
Just updated with new positions, now at 70 companies actively hiring in Athens, Thessaloniki, Crete, Ioannina, and remote
Some really cool opportunities in there, from vertical AI to space tech and more
startuppirate.com/job-board
Christos Baziotis retweeted
Every developer has an IDE. Researchers never had one.
We don’t need another research agent, but a full stack workstation powered by AI.
Today we're launching Orchestra @orch_research, the world's first Research IDE 🔬🧵
Christos Baziotis retweeted
@cohere just released the best speech->text model :)
It currently ranks #1 for accuracy on @huggingface Open ASR Leaderboard, setting a new benchmark for real-world transcription performance. Read more 👇
Christos Baziotis retweeted
Last week, Sierra released 𝜏³-Bench, an update to our industry standard agent benchmark, expanding the benchmark in several important ways:
𝜏-Knowledge tests whether agents can operate over large collections of internal company documents spread across systems and formats;
𝜏-Voice evaluates agents built for live voice conversations; and
We’ve also incorporated fixes contributed by the 𝜏-Bench community across existing 𝜏-Bench domains to improve evaluation accuracy.
Read more here: sierra.ai/blog/bench-advanci…
Christos Baziotis retweeted
We're back with a new website, hundreds of job openings at startups in Greece, and a list of 65+ teams hiring right now
Browse roles in product, engineering, sales, HR, and more: startuppirate.com/job-board
Christos Baziotis retweeted
seems big AI labs are hyperfixating on reasoning when they should focus on *memory* instead
normal people won't use models that can think for hours to solve hard math problems
people want models that learn over time, remember details, adapt and interact like a person would
Christos Baziotis retweeted
Is Agentic RAG worth it? An experimental comparison of RAG approaches
Compares Enhanced RAG (fixed pipelines with modules like rerankers) vs Agentic RAG across multiple dimensions, finding neither universally superior but Agentic costs up to 3.6x more.
📝arxiv.org/abs/2601.07711
Christos Baziotis retweeted
Introducing our latest breakthrough in AI search and retrieval: Rerank 4!
It’s the most advanced set of reranking models on the market, with best-in-class performance across search relevance, speed, deployment flexibility, multilingual support, and domain-specific understanding.
Christos Baziotis retweeted
📚Tom Sherborne: Modeling Cross-lingual Transfer for Semantic Parsing
Sherborne’s dissertation developsmethods for cross-lingual transfer into low-resource languages, demonstrating their effectiveness in the context of semantic parsing for integration with database APIs.
Christos Baziotis retweeted
Even great answers can have subtle gaps—missing numbers, overlooked facts. In high-stakes workflows, those gaps cost time and trust.
At Samaya, we built a real-time self-correction engine that quality-checks our summaries and fills in missing info before the user notices.
🧵👇
samaya.ai/blog/spotting-gaps…
Christos Baziotis retweeted
Wait for 9 minutes or get it instantly and in more depth on @samaya_AI
DMs open if you want access!
Christos Baziotis retweeted
Evaluating long-form answers to complex technical questions is very challenging.
Existing methods fall short in this setting.
At Samaya, we built Criteria-Eval, a checklist-based evaluation that aligns with how domain experts judge answers.
🧵samaya.ai/blog/criteria-eval
✍️ @cbaziotis
Christos Baziotis retweeted
P.S. It turns out the fact that RL can only amplify weak signals has been known to the NMT community for a long time.
nitter.gq/kchonyc/status/1927435…
Christos Baziotis retweeted
son of a bitch!!!!
**ALL** of my GRPO runs have been done with an extremely narrow subset of the probability distribution!!!
why is it even fallbacking to that in the first place if vllm accepts `topk = -1` just fine!!!!
