Sima Noorani (@NooraniSimaa) - โปรไฟล์ Twitter

ทวีตที่ปักหมุด

Sima Noorani@NooraniSimaa·25 Kas

When humans and AI collaborate, what should uncertainty quantification look like? Our new paper proposes two principles---no counterfactual harm and complementarity---and gives distribution-free guarantees without assumptions on the task, AI model, or human behavior.

English

4

10

99

9.5K

Sima Noorani รีทวีตแล้ว

Shayan Kiyani@ShayanKiyani1·1d

If reliable verification is expensive, then intelligence is partly the art of spending it wisely. Reasoning with LLMs increasingly unfolds through weak (fast but noisy) and strong (reliable but costly) verification. The question is: how should we orchestrate the two?

English

1

6

22

1.2K

Sima Noorani@NooraniSimaa·10 Mar

Paper:arxiv.org/pdf/2602.17646 Joint work with the amazing @ShayanKiyani1, @HamedSHassani, @pappasg69

English

0

1

6

275

Sima Noorani@NooraniSimaa·10 Mar

Empirically, we study the framework in both LLM-simulated interactions and a real human crowdsourcing study, and show that enforcing these two principles leads to predictable shifts in downstream human decision quality. This validates that these human-centric principles serve as practical levers for steering multi-round HAI collaboration dynamics.

English

1

0

146

Sima Noorani@NooraniSimaa·10 Mar

In real-world multi-round human–AI interactions, the human decides whether to engage, how long to continue, and how much to trust the AI, while bearing accountability for the final outcome. How should these interactions be designed to incentivize human engagement in a principled way?

English

1

2

14

445

Sima Noorani@NooraniSimaa·3 Ara

We’re presenting our work today at Poster Session 2 (4:30 PM, #3303). Come check it out and chat with us! :) This is a joint work with @ShayanKiyani1 @HamedSHassani @pappasg69.

Sima Noorani@NooraniSimaa

How can we quantify uncertainty in LLMs from only a few sampled outputs? The key lies in the classical problem of missing mass—the probability of unseen outputs. This perspective offers a principled foundation for conformal prediction in query-only settings like LLMs.

English

1

3

30

2.1K

Sima Noorani@NooraniSimaa·25 Kas

Paper : arxiv.org/pdf/2510.23476 Joint work with the amazing @ShayanKiyani1 , @pappasg69, @HamedSHassani

English

1

0

7

552

Sima Noorani@NooraniSimaa·25 Kas

Empirically, the resulting collaborative sets improve over both human-only and AI-only sets in both marginal coverage and average set size.

English

1

0

2

599

Sima Noorani@NooraniSimaa·25 Kas

When humans and AI collaborate, what should uncertainty quantification look like? Our new paper proposes two principles---no counterfactual harm and complementarity---and gives distribution-free guarantees without assumptions on the task, AI model, or human behavior.

English

4

10

99

9.5K

Sima Noorani รีทวีตแล้ว

Aaron Roth@Aaroth·30 Eki

How should you use forecasts f:X->R^d to make decisions? It depends what properties they have. If they are fully calibrated (E[y | f(x) = p] = p), then you should be maximally agressive and act as if they are correct --- i.e. play argmax_a E_{o ~ f(x)}[u(a,o)]. On the other hand

English

1

14

98

8.8K

Sima Noorani รีทวีตแล้ว

Shayan Kiyani@ShayanKiyani1·12 Haz

We push conformal prediction and its trade-offs beyond regression & classification — into query-based generative models. Surprisingly (or not?), missing mass & Good-Turing estimators emerge as key tools once again. Very excited about this one!

Sima Noorani@NooraniSimaa

How can we quantify uncertainty in LLMs from only a few sampled outputs? The key lies in the classical problem of missing mass—the probability of unseen outputs. This perspective offers a principled foundation for conformal prediction in query-only settings like LLMs.

English

0

4

24

2.8K

Sima Noorani

ค้นพบ