Archiki Prasad
@ArchikiPrasad
Followers
1K
Following
2K
Statuses
353
PhD student @uncnlp, advised by @mohitban47 | Intern: @AIatMeta (FAIR) | Prev: @allenai_org @AdobeResearch; Research interests: #NLProc #ML
Chapel Hill, NC
Joined December 2016
🚨 Excited to share: "Learning to Generate Unit Tests for Automated Debugging" 🚨 which introduces ✨UTGen and UTDebug✨ for teaching LLMs to generate unit tests (UTs) and debugging code from generated tests. UTGen+UTDebug improve LLM-based code debugging by addressing 3 key questions: 1⃣ What are desirable properties of unit test generators? (A: high output acc and rate of uncovering errors) 2⃣ How good are models at 0-shot unit test generation (A: they are not great) ... so how do we improve LLMs' UT generation abilities? (A: bootstrapping from code-generation data via UTGen) 3⃣ How can we use potentially noisy feedback from generated tests for debugging? (A: via test-time scaling and validation + backtracking in UTDebug) 🧵👇
4
58
163
@rohanpaul_ai Thanks for sharing our work! For folks interested in more details, check out our thread 🧵⬇️
🚨 Excited to share: "Learning to Generate Unit Tests for Automated Debugging" 🚨 which introduces ✨UTGen and UTDebug✨ for teaching LLMs to generate unit tests (UTs) and debugging code from generated tests. UTGen+UTDebug improve LLM-based code debugging by addressing 3 key questions: 1⃣ What are desirable properties of unit test generators? (A: high output acc and rate of uncovering errors) 2⃣ How good are models at 0-shot unit test generation (A: they are not great) ... so how do we improve LLMs' UT generation abilities? (A: bootstrapping from code-generation data via UTGen) 3⃣ How can we use potentially noisy feedback from generated tests for debugging? (A: via test-time scaling and validation + backtracking in UTDebug) 🧵👇
1
0
2
RT @rohanpaul_ai: The challenge lies in effectively debugging faulty code produced by LLMs due to the scarcity of unit tests that can pinpo…
0
8
0
RT @ArchikiPrasad: 🚨 Excited to share: "Learning to Generate Unit Tests for Automated Debugging" 🚨 which introduces ✨UTGen and UTDebug✨ for…
0
58
0
RT @mohitban47: 🚨 Check out "UTGen & UTDebug" for learning to automatically generate unit tests (i.e., discovering inputs which break your…
0
18
0
RT @kuchaev: Our team put together a unified mathematical framework to analyze popular model alignment algorithms. “Reward-aware Preference…
0
19
0
RT @ADarmouni: Models can be finetuned to generate adversarial unit tests for debugging 📖 Read of the day, season 3, day 13: « Learning to…
0
1
0
FYI: The debugging benchmarks we developed in our work to evaluate unit test generators & debuggers are now available on @huggingface: We construct multiple datasets with subtly flawed code ranging in debugging difficulty. More details 🧵⬇️
🚨 Excited to share: "Learning to Generate Unit Tests for Automated Debugging" 🚨 which introduces ✨UTGen and UTDebug✨ for teaching LLMs to generate unit tests (UTs) and debugging code from generated tests. UTGen+UTDebug improve LLM-based code debugging by addressing 3 key questions: 1⃣ What are desirable properties of unit test generators? (A: high output acc and rate of uncovering errors) 2⃣ How good are models at 0-shot unit test generation (A: they are not great) ... so how do we improve LLMs' UT generation abilities? (A: bootstrapping from code-generation data via UTGen) 3⃣ How can we use potentially noisy feedback from generated tests for debugging? (A: via test-time scaling and validation + backtracking in UTDebug) 🧵👇
0
11
41
RT @codezakh: Testing is a critical part of software engineering — what if we could automatically discover inputs which break your code? T…
0
6
0
RT @cyjustinchen: Introducing ✨UTGen & UTDebug✨ for improving code debugging tasks w/ strong pass@1 gains. Unit tests help both humans and…
0
8
0
RT @EliasEskin: 🚨 Excited to announce UTGen and UTDebug, where we first learn to generate unit tests and then apply them to debugging gener…
0
10
0
Thanks to my amazing co-authors @EliasEskin (co-lead), @cyjustinchen , @codezakh and @mohitban47 for a great collaboration! @unccs, @uncnlp Code+Datasets: Paper:
0
2
11
RT @jaseweston: 🚨 Diverse Preference Optimization (DivPO) 🚨 SOTA LLMs have model collapse🫠: they can't generate diverse creative writing or…
0
79
0
RT @hanlin_hl: Happy to share that "VEDiT: Latent Prediction Architecture For Procedural Video Representation Learning" has been accepted t…
0
23
0