@ConorFinlay3

Research scholar @cohere

 London
linkedin.com/in/conorfinlay/
Joined August 2019
  • Tweets18
  • Following401
  • Followers30
5Media
I'll be here to present our team's work on confidence estimation in reasoning models, come by our booth!
Cohere Labs
@Cohere_LabsOct 16
Cohere Labs and @cohere are proud to share some of our latest research next week at EMNLP 2026, in Budapest, Hungary! 🇭🇺 Very honored to be featured as part of this conference focused on cutting-edge deep learning research and its real-world applications.
2
5
364
Conor Finlay retweeted
Cohere Labs
@Cohere_LabsOct 14
Announcing new research on scaling multi-agent teams for long horizon software engineering tasks. 🤖🧩📈 Work led by @keenansamway, @mrdanieldsouza, Omar Darwiche Domingues, @mziizm, @yichern_tan, and @nouhadziri Read the paper: cohere.link/tStRYJS Check out the blog post: cohere.link/0TzgKoy
12
40
14
368
39,377
Conor Finlay retweeted
Sebastian Raschka
@rasbtJun 26
Have been taking different local open-weight LLMs for a test drive in different harnesses (Qwen-Code, Codex, Claude Code). 30B Mixture-of-Expert models are kind of a nice sweet spot and can solve challenging problems. And they get roughly 40 tok/sec on a Mac or DGX Spark, which is similar to GPT 5.5 in a Pro subscription and totally useable for everyday work. More interesting is also the harness choice! Claude Code seems to be using 2x many tokens as Codex. Gemma 4 E2B is here just for reference to show that the tasks can't be trivially solved by smaller models. Just finishing a longer write-up about this and will share soon (likely tomorrow)!
117
104
11
901
77,760
Conor Finlay retweeted
Cohere Labs
@Cohere_LabsJun 24
Announcing new research on calibrating confidence in reasoning models 🎯⚖️, led by @ConorFinlay3, @joshuakurien_, @TheyCallMeMr_, @mziizm, and @beyzaermis. 📜Read the paper: arxiv.org/abs/2606.24281
1
6
1
25
4,805
Can an LLM know when not to trust its own answer? 🤔 Reasoning models are powerful but overconfident, and their verbalised confidence is generally unreliable. We built CALIBER: an RL method that teaches a model to estimate its confidence twice, before and after reasoning.
1
3
11
1,180
more replies
A nice side effect: eliciting confidence before and after reasoning also shapes how the model reasons. You can watch it start uncertain, work through what it knows, and arrive at a confident answer. 👇
1
1
3
158
Read the full paper here: arxiv.org/abs/2606.24281 Work with @joshuakurien_, @TheyCallMeMr_, @mziizm, and @beyzaermis.

CALIBER: Calibrating Confidence Before and After Reasoning in...

Reasoning language models are increasingly asked not only to answer difficult questions, but also to estimate their likelihood of success. Existing methods typically elicit confidence only once:...

arxiv.org
1
2
182
Conor Finlay retweeted
Cohere
@cohereJun 9
Introducing Cohere's first open-source coding model: North Mini Code Small & efficient, designed for agentic performance and built for community input.

Meet North Mini Code

Cohere's first Apache 2.0 coding model

68
263
101
2,268
600,918
Conor Finlay retweeted
Aidan Gomez
@aidangomezJun 15
Excited to triple down on London! 🇬🇧 thetimes.com/business/techno…

Canada’s Cohere expands in Britain to cash in interest on AI

The artificial intelligence start-up has tripled its UK footprint as it looks to be an alternative to US competitors OpenAI and Anthropic

thetimes.com
14
15
4
263
36,797
Conor Finlay retweeted
Cohere
@cohereMar 26
Introducing: Cohere Transcribe – a new state-of-the-art in open source speech recognition.
81
291
86
2,544
613,459
Conor Finlay retweeted
Goodfire
@GoodfireAI6 Aug 2025
The activation space trajectories of two instances of gpt-oss talking to each other seem to converge on a couple attractors (in blue). Red is a control (Claude as interlocutor, pretending to be a normal chatbot user).
4
2
39
2,216
Load more