Twitxy Saved

Eshwar Chandrasekharan

@eshwar_chan

United States eshwarchandrasekharan.com Joined September 2011
106 Following 607 Followers 195 Posts

Latest posts 5

Eshwar Chandrasekharan @eshwar_chan
The words you pick online aren't just style choices. They shape which ideas get seen and rewarded. And most community guidelines don't even address this.

Excited to see @_agam_goyal_ present this work at #CHI2026!
Agam Goyal Verified account @_agam_goyal_ Apr 8
🚨 New paper alert at #CHI2026!

In order to learn from aggregated community preferences, we often treat upvotes, golds, and awards as signals of what communities value. But which linguistic features actually drive those signals, and which are just surface correlations?

I’ll be presenting my work "The Language of Approval: Identifying the Drivers of Positive Feedback Online" at CHI 2026 in Barcelona, Spain 🇪🇸
🔗 arxiv.org/pdf/2509.10370

In this paper, we study 100 Reddit communities and use causal inference to identify which linguistic features actually drive positive feedback like upvotes, gold, and awards.

Huge thanks to @lambert_char and my advisor @eshwar_chan for their help!

More details in 🧵 [1/7]
View quoted post on X
Eshwar Chandrasekharan @eshwar_chan
Q: Are fine-tuned SLMs better than LLMs at identifying online content that violate community norms?🤖⚖️

A: Yes!

Find out more in our #NAACL2025 paper co-lead by @_agam_goyal_ and Xianyang Zhan, in collaboration with Yilun Chen and @kous2v

📄 arxiv.org/pdf/2410.13155
Agam Goyal Verified account @_agam_goyal_ Apr 25, 2025
🚨 New #NAACL2025 paper alert: “SLM-Mod: Small Language Models Surpass LLMs at Content Moderation”

We show that fine-tuned small language models outperform LLMs like GPT-4o for content moderation tasks, with higher accuracy and recall!

🔗 arXiv: arxiv.org/pdf/2410.13155

🧵[1/7]
View quoted post on X
0 replies 1 reposts 8 likes 1.3K views View on X