Hadas Orgad

Research Fellow @ Kempner Institute, Harvard | Working on AI interpretability, robustness & safety

[Hadas Orgad @OrgadHadas]

I'm excited to share that I'll be joining @KempnerInst @Harvard as a research fellow this September!

[Kempner Institute at Harvard University @KempnerInst]

Thrilled to announce the 2025 recipients of #KempnerInstitute Research Fellowships: Elom Amemastro, Ruojin Cai, David Clark, Alexandru Damian, William Dorrell, Mark Goldstein, Richard Hakim, Hadas Orgad, Gizem Ozdil, Gabriel Poesia, & Greta Tuckute!

https://t.co/mbFOsLFdNw

[Hadas Orgad @OrgadHadas]

@SohamPahari @ActInterp Thanks! We'll fix it. It's 5 pages

[Hadas Orgad @OrgadHadas]

Submit your work! The 2nd Workshop on 𝐀𝐜𝐭𝐢𝐨𝐧𝐚𝐛𝐥𝐞 𝐈𝐧𝐭𝐞𝐫𝐩𝐫𝐞𝐭𝐚𝐛𝐢𝐥𝐢𝐭𝐲 will be held at COLM 2026 in San Francisco!

Submission Deadline: June 21, 2026

@ActInterp

[Hadas Orgad @OrgadHadas]

@ActInterp CFP is here >> https://t.co/EeYxNyyteN

[Hadas Orgad @OrgadHadas]

@eb1aexperts Thank you!

[Hadas Orgad @OrgadHadas]

Excited that our paper on Actionable Interpretability got accepted to ICML!
And just in time -- we also heard that our Actionable Interpretability workshop will be happening again, in COLM!
See you in Korea 🇰🇷 and SF🌉
[Arxiv paper link in the comment]

[Hadas Orgad @OrgadHadas]

Our ICML 2025 workshop on Actionable Interpretability drew massive interest. But the same questions kept coming up: What does "actionable" mean? Is it achievable? How?
We're ready to answer.
🧵

[Hadas Orgad @OrgadHadas]

Arxiv >> https://t.co/3M8q8GD6Km
Blog post >> https://t.co/5dbTegfTZG

[Yanai Elazar @yanaiela]

People seem to really freak out about hallucinated citations as the "bad consequence of AI slop" but
(1) it's easy to detect (and fix), and
(2) it's so insignificant compared to other erroneous/bad/misleading writing AI can make in scientific papers.

[Hadas Orgad @OrgadHadas]

@naghmehfarzi @TamarRottShaham @MIT_CSAIL @KempnerInst The talk was not recorder, but I gave a similar talk on the same work at Mila tea talks in February > https://t.co/64sCOEbMSv

[Hadas Orgad @OrgadHadas]

@DataSciNews @TamarRottShaham @MIT_CSAIL @KempnerInst Legs ;)

[Hadas Orgad @OrgadHadas]

???

Spotted in SF

[Hadas Orgad @OrgadHadas]

@DifanJ2000 Do you think that the generalization is related to your feature choice? E.g., did you test generalization on a "vanilla" layer-wise linear probe?

[Stanford NLP Group @stanfordnlp]

For this week's NLP seminar, we are excited to host @OrgadHadas from Harvard University!

Date and Time: Thursday, April 30, 11:00 AM — 12:00 PM Pacific Time.
Zoom Link: https://t.co/jmz2wb8pIP

Title: Large Language Models Generate Harmful Content Using a Distinct, Unified Mechanism

Abstract: We use targeted weight pruning as a causal intervention to probe the internal organization of harmfulness in LLMs. Our results reveal a coherent internal structure for harmfulness in LLMs that may serve as a foundation for more principled approaches to safety.

Hope to see you all there!

[Hadas Orgad @OrgadHadas]

@SimonSchrodi Let's chat today!

[Hadas Orgad @OrgadHadas]

In this work, led by Joe, we evaluate a wide range of truthfulness probes and show they *still* fail to robustly generalize.

We draw lessons for how these probes should be evaluated, and identify design choices that can improve robustness.

[Joe Stacey @_joestacey_]

Excited to share my first postdoc paper with @SheffieldNLP ! 🤩

In this work we argue that supervised uncertainty quantification (UQ) needs better evaluation

Want to know more? Here's a little summary 🧵

[Hadas Orgad @OrgadHadas]

@seraphinagt Paper is here, fresh out of the Overleaf
https://t.co/4JhUx3jvyl

[Hadas Orgad @OrgadHadas]

@FazlBarez Apparently me too! I wasn't aware. About half of these are inappropriate messages from bots 🤔

[Hadas Orgad @OrgadHadas]

@dmhook Is this SNIP plus the steering or just SNIP? What does, for example, 98% compliance and 100% safe coherence mean?

The refusal ablation seems to work best with p =~ q. Through all of my experiments, p=q=0.01 worked best (highest harmfulness).

[Hadas Orgad @OrgadHadas]

New paper: LLMs encode harmful content generation in a distinct, unified mechanism

Using weight pruning, we find that harmful generation depends on a tiny subset of the weights that are shared across harm types and separate from benign capabilities.

🧵

[Hadas Orgad @OrgadHadas]

@ASM65617010 @davidad More distributed is one hypothesis. Within humans, I don't think we know that it's distributed. We don't have such percise intervention in neuroscience.

[Hadas Orgad @OrgadHadas]

@OwainEvans_UK @BetleyJan This now inspired us to run more experiments to explain the difference between models. We'll share what we find.

[Hadas Orgad @OrgadHadas]

@OwainEvans_UK @BetleyJan It’s a good point that the removal isn’t as complete in qwen, which opens up some interesting questions for follow up. For example, is EM more entangled with other basic pretrained concepts because qwen is exposed to instruction data in pretraining?