iiinova — where your digital self keeps evolving

Reposted by @aakriti1kumar.bsky.social

Matt Groh @mattgroh.bsky.social · 11/02/2026

🚨 New in @natmachintell.nature.com 🚨 We collected 9000+ annotations of empathic communication in convos from experts, crowds & LLMs across 4 NLP/comms/psych frameworks LLM judgment exceeds crowds' reliability & nearly matches experts Soft skills can now be reliably measured by LLMs 🧵

2144

aakriti1kumar.bsky.social @aakriti1kumar.bsky.social · 17/06/2025

There’s a lot more detail in the full paper, and I would love to hear your thoughts and feedback on it! Check out the preprint here: arxiv.org/pdf/2506.10150

arxiv.org

010

aakriti1kumar.bsky.social @aakriti1kumar.bsky.social · 17/06/2025

LLMs, when benchmarked against reliable expert judgments, can be reliable tools for overseeing emotionally sensitive AI applications. Our results show we can use LLMs-as-judge to monitor LLMs-as-companion!

110

aakriti1kumar.bsky.social @aakriti1kumar.bsky.social · 17/06/2025

For example, in one of the conversations in our dataset, a response that an expert saw as "dismissing” the speaker’s emotions, a crowdworker interpreted as "validating" their emotions instead!

110

aakriti1kumar.bsky.social @aakriti1kumar.bsky.social · 17/06/2025

These misjudgments from crowdworkers have huge implications for AI training and deployment❌ If we use flawed evaluations to train and monitor "empathic" AI, we risk creating systems that propagate a broken standard of what good communication looks like.

110

aakriti1kumar.bsky.social @aakriti1kumar.bsky.social · 17/06/2025

So why the gap between experts/LLMs and crowds? Crowdworkers often - have limited attention - rely on heuristics like “it’s the thought that counts” - focusing on intentions rather than actual wording show systematic rating inflation due to social desirability bias

120

aakriti1kumar.bsky.social @aakriti1kumar.bsky.social · 17/06/2025

And when experts disagree, LLMs struggle to find a consistent signal too. Here’s how expert agreement (Krippendorff's alpha) varied across empathy sub-components:

110

aakriti1kumar.bsky.social @aakriti1kumar.bsky.social · 17/06/2025

But here’s the catch: LLMs are reliable when experts are reliable. The reliability of expert judgments depends on the clarity of the construct. For nuanced, subjective components of empathic communication, experts often disagree.

110

aakriti1kumar.bsky.social @aakriti1kumar.bsky.social · 17/06/2025

We analyzed thousands of annotations from LLMs, crowdworkers, and experts on 200 real-world conversations And specifically looked at 21 sub-components of empathic communication from 4 evaluative frameworks The result? LLMs consistently matched expert judgments better than crowdworkers did! 🔥

110

aakriti1kumar.bsky.social @aakriti1kumar.bsky.social · 17/06/2025

How do we reliably judge if AI companions are performing well on subjective, context-dependent, and deeply human tasks? 🤖 Excited to share the first paper from my postdoc (!!) investigating when LLMs are reliable judges - with empathic communication as a case study 🧐 🧵👇

140

aakriti1kumar.bsky.social @aakriti1kumar.bsky.social · 02/04/2025

Super cool opportunity to work with brilliant scientists and fantastic mentors @mattgroh.bsky.social and Dashun Wang 🌟🌟 Feel free to reach out!

020

Reposted by @aakriti1kumar.bsky.social

Abhishek Sharma @abhishekshar.bsky.social · 23/01/2025

Our paper: Decision-Point Guided Safe Policy Improvement We show that a simple approach to learn safe RL policies can outperform most offline RL methods. (+theoretical guarantees!) How? Just allow the state-actions that have been seen enough times! 🤯 arxiv.org/abs/2410.09361

arxiv.org

Decision-Point Guided Safe Policy Improvement

Within batch reinforcement learning, safe policy improvement (SPI) seeks to ensure that the learnt policy performs at least as well as the behavior policy that generated the dataset. The core challeng...

041