Kshitish Ghate

58 posts

Kshitish Ghate

@GhateKshitish

PhD student @UWCSE | MLT Grad student @LTIatCMU | CS and Econ @bitspilanigoa

Katılım Ekim 2022

253 Takip Edilen100 Takipçiler

Sabitlenmiş Tweet

Kshitish Ghate@GhateKshitish·14 Eki

🚨New paper: Reward Models (RMs) are used to align LLMs, but can they be steered toward user-specific value/style preferences? With EVALUESTEER, we find even the best RMs we tested exhibit their own value/style biases, and are unable to align with a user >25% of the time. 🧵

English

6.5K

Kshitish Ghate retweetledi

Natasha Jaques@natashajaques·3d

The paper I’ve been most obsessed with lately is finally out: nbcnews.com/tech/tech-news…! Check out this beautiful plot: it shows how much LLMs distort human writing when making edits, compared to how humans would revise the same content. We take a dataset of human-written essays from 2021, before the release of ChatGPT. We compare how people revise draft v1 -> v2 given expert feedback, with how an LLM revises the same v1 given the same feedback. This enables a counterfactual comparison: how much does the LLM alter the essay compared to what the human was originally intending to write? We find LLMs consistently induce massive distortions, even changing the actual meaning and conclusions argued for.

English

385

1.4K

237.2K

Kshitish Ghate retweetledi

Zora Wang@ZhiruoW·11 Şub

‼️Position: AI coding agent research needs recalibration. We've heavily optimized for solo autonomy, and far less for designing agents that empower the humans using them. It’s time to build human-centered coding agents. 🧵

English

317

48.7K

Kshitish Ghate retweetledi

Andy Liu@uilydna·5 Şub

Our work eliciting which values LLMs prioritize via simulated value conflicts was accepted to #ICLR2026 ! See you in🇧🇷

Andy Liu@uilydna

🚨New Paper: LLM developers aim to align models with values like helpfulness or harmlessness. But when these conflict, which values do models choose to support? We introduce ConflictScope, a fully-automated evaluation pipeline that reveals how models rank values under conflict.

English

4.6K

Kshitish Ghate retweetledi

Ethan Shen@ethnlshn·27 Oca

Today, we release SERA-32B, an approach to coding agents that matches Devstral 2 at just $9,000. It is fully open-source and you can train your own model easily - at 26x the efficiency of using RL. Paper: allenai.org/papers/opencod… Here’s how 🧵

Ai2@allen_ai

Introducing Ai2 Open Coding Agents—starting with SERA, our first-ever coding models. Fast, accessible agents (8B–32B) that adapt to any repo, including private codebases. Train a powerful specialized agent for as little as ~$400, & it works with Claude Code out of the box. 🧵

English

691

90.3K

Kshitish Ghate retweetledi

Andy Liu@uilydna·24 Eki

@jifan_zhang @johnschulman2 @sleight_henry @TheAndiPenguin Congrats on the paper! We've been working in a similar direction (evaluating value prioritization in LLMs - arxiv.org/abs/2509.25369), would love to chat if you're still thinking about this

English

2.1K

Kshitish Ghate retweetledi

Lucy Li@lucy3_li·16 Eki

PhD apps season is here! 😱🥳 Apply to do a PhD @WisconsinCS (as pictured) w/ me to research: - Societal impact of AI - NLP ←→ CSS and cultural analytics - Computational sociolinguistics - Human-AI interaction - Culturally competent and inclusive NLP lucy3.github.io/prospective-st…

English

362

57.6K

Kshitish Ghate retweetledi

Taylor Sorensen@ma_tay_·15 Eki

My best hypothesis for the mechanism is: Chat LLMs are hyperoptimized to approximate the single "best" (most-preferred) response. When you prompt it for a single story, it gives the single best story it can. When you ask it to give FIVE stories, you recast the "best" response to be one containing FIVE stories, which has more diversity (a very good trick!) However, in the limit, as we train models with this objective, it converges to ALWAYS giving the same "best"/high-reward story - a fundamental limitation of the current paradigm x.com/ma_tay_/status…

English

1.8K

Kshitish Ghate@GhateKshitish·14 Eki

Work done with amazing collaborators 🙏 @uilydna @devanshrjain @ma_tay_ @Dr_Atoosa @aylin_cim @MonaDiab77 @MaartenSap

English

510

Kshitish Ghate@GhateKshitish·14 Eki

For more details about our experiments and findings -- Paper: arxiv.org/abs/2510.06370 Code and Data: github.com/kshitishghate/… Please feel free to reach out if you are interested in this work and would like to chat!

English

218

Kshitish Ghate@GhateKshitish·14 Eki

English

6.5K

Kshitish Ghate retweetledi

Taylor Sorensen@ma_tay_·13 Eki

🤖➡️📉 Post-training made LLMs better at chat and reasoning—but worse at distributional alignment, diversity, and sometimes even steering(!) We measure this with our new resource (Spectrum Suite) and introduce Spectrum Tuning (method) to bring them back into our models! 🌈 1/🧵

English

198

68K

Kshitish Ghate@GhateKshitish·2 Eki

Check out our new paper that uses simulated moral dilemmas to study how LLMs prioritize different values!

Andy Liu@uilydna

English

168

Keşfet

@jifan_zhang @johnschulman2 @sleight_henry @TheAndiPenguin @WisconsinCS @uilydna @devanshrjain @ma_tay_