Lightnews — Scholar-powered news

LightNews

Lakshya A Agrawal

@lakshyaaagrawal.bsky.social

760 followers 3.5K following 35 posts

PhD @ucberkeleyofficial.bsky.social | Past: AI4Code Research Fellow @msftresearch.bsky.social | Summer @EPFL Scholar, CS and Applied Maths @IIITDelhi | Hobbyist Saxophonist https://lakshyaaagrawal.github.io Maintainer of https://aka.ms/multilspy

lakshyaaagrawal.github.io

Posts Media Videos Starter Packs

Reposted by Lakshya A Agrawal

Nico Halecky @nehalecky.bsky.social · 12d

Just what I was looking for. Thank you for sharing, looking forward to the read.

Reposted by Lakshya A Agrawal

Sung Kim @sungkim.bsky.social · 12d

propose and test prompt updates, and combine complementary lessons from the Pareto frontier of its own attempts.

arxiv.org/abs/2507.19457

GEPA: Reflective Prompt Evolution Can Outperform Reinforcement Learning

Large language models (LLMs) are increasingly adapted to downstream tasks via reinforcement learning (RL) methods like Group Relative Policy Optimization (GRPO), which often require thousands of rollo...

Reposted by Lakshya A Agrawal

Sung Kim @sungkim.bsky.social · 12d

DSPy folks love GEPA, so here's a GEPA paper for anyone who wants to learn more.

Given any AI system containing one or more LLM prompts, GEPA samples system-level trajectories (e.g., reasoning, tool calls, and tool outputs) and reflects on them in natural language to diagnose problems,

Reposted by Lakshya A Agrawal

Me AI @tbressers.bsky.social · 12d

..GEPA and prompt optimization explained: https://arxiv.org/abs/2507.19457v1

(7/7)

ArXiv page 7

Reposted by Lakshya A Agrawal

Me AI @tbressers.bsky.social · 12d

..make adapting large models more practical—especially when compute or data is limited. It’s like giving AI a way to learn from its own “thinking out loud,” turning natural language into a powerful tool for self-improvement.

Links:
Paper on arXiv: https://arxiv.org/abs/2507.19457 ..

(6/7)

ArXiv page 6

Reposted by Lakshya A Agrawal

Me AI @tbressers.bsky.social · 12d

..code on the fly.

What’s cool here is the shift from treating AI tuning as a blind search for a higher score to a reflective process that leverages the AI’s native strength: language. By evolving prompts through thoughtful reflections, GEPA unlocks smarter, faster learning that could..

(5/7)

ArXiv page 5

Reposted by Lakshya A Agrawal

Me AI @tbressers.bsky.social · 12d

..fewer attempts than traditional reinforcement learning methods. On several tough tasks like multi-step question answering and instruction following, GEPA consistently outperforms both standard reinforcement learning and previous prompt optimizers. It even shows promise for optimizing..

(4/7)

ArXiv page 4

Reposted by Lakshya A Agrawal

Me AI @tbressers.bsky.social · 12d

..strategies by mixing and matching what works best.

GEPA treats AI prompt tuning like a conversation with itself, iterating through generations of prompts that learn from detailed feedback written in words, not just numbers. This lets it learn much more efficiently—up to 35 times..

(3/7)

ArXiv page 3

Reposted by Lakshya A Agrawal

Me AI @tbressers.bsky.social · 12d

..what went wrong and how to fix it? That’s the idea behind a new approach called GEPA. Instead of relying solely on those sparse reward signals, GEPA has AI inspect its own attempts using natural language reflections. It diagnoses errors, proposes prompt fixes, and evolves smarter..

(2/7)

ArXiv page 2

Reposted by Lakshya A Agrawal

Me AI @tbressers.bsky.social · 12d

What if language itself could teach AI to get better, faster?

Most AI training feels like trial and error in the dark—reinforcement learning tweaks models by chasing a number, often needing tens of thousands of tries to improve. But what if the AI could actually *talk to itself* about..

(1/7)

ArXiv page 1

Reposted by Lakshya A Agrawal

LLMs @llms.activitypub.awakari.com.ap.brid.gy · 16d

gepa 0.0.15a1 A framework for optimizing textual system components (AI prompts, code snippets, etc.) using LLM-based reflection and Pareto-efficient evolutionary search.

Origin | Interest | Match

A framework for optimizing textual system components (AI prompts, code snippets, etc.) using LLM-based reflection and Pareto-efficient evolutionary search.

Reposted by Lakshya A Agrawal

Awakari @bluesky.awakari.com · 16d

gepa 0.0.15a1 A framework for optimizing textual system components (AI prompts, code snippets, etc.) using LLM-based reflection and Pareto-efficient evolutionary search.

Interest | Match | Feed

Reposted by Lakshya A Agrawal

Awakari @bluesky.awakari.com · 16d

gepa 0.0.15a1 A framework for optimizing textual system components (AI prompts, code snippets, etc.) using LLM-based reflection and Pareto-efficient evolutionary search.

Interest | Match | Feed

Reposted by Lakshya A Agrawal

Awakari @bluesky.awakari.com · 16d

gepa 0.0.16 A framework for optimizing textual system components (AI prompts, code snippets, etc.) using LLM-based reflection and Pareto-efficient evolutionary search.

Interest | Match | Feed

Reposted by Lakshya A Agrawal

Awakari @bluesky.awakari.com · 16d

gepa 0.0.16 A framework for optimizing textual system components (AI prompts, code snippets, etc.) using LLM-based reflection and Pareto-efficient evolutionary search.

Interest | Match | Feed

Reposted by Lakshya A Agrawal

Sean Michael Kerner @techjournalist.bsky.social · 14d

New research released today from Databricks shows how its GEPA (Generative Evolutionary Prompt Adaptation) technique improves prompt optimization by an order of magnitude.

venturebeat.com/ai/the-usd10...

venturebeat.com

Reposted by Lakshya A Agrawal

qdrddr.bsky.social @qdrddr.bsky.social · 13d

🚀 #GEPA: Automatic #Prompt Optimization by @databricksinc.bsky.social: gpt-oss-120b beats Claude Sonnet 4 (+3%) at ~20x lower cost. Completes with DSPy SIMBA/MIPROv2
📜 MIT lic
🔗 Link in first 💬⤵️

Repost 🔁 #AI #LLM #RAG #PromptEngineering #ContextEngineering

Reposted by Lakshya A Agrawal

qdrddr.bsky.social @qdrddr.bsky.social · 13d

⭐️🔗 GitHub: github.com/gepa-ai/gepa
⭐️🔗 www.databricks.com/blog/buildin...
👉 Discuss with me in Discord: linktr.ee/qdrddr

GitHub - gepa-ai/gepa: Optimize prompts, code, and more with AI-powered Reflective Text Evolution

Optimize prompts, code, and more with AI-powered Reflective Text Evolution - gepa-ai/gepa

Reposted by Lakshya A Agrawal

Bilal Tariq @bilaltariq01.bsky.social · 28d

Lessons & Practices for Building and Optimizing Multi-Agent RAG Systems with DSPy and GEPA https://cstu.io/cbe4bb #love #techno #future

Lessons & Practices for Building and Optimizing Multi-Agent RAG Systems with DSPy and GEPA

Introduction When I first read “Building and Optimizing Multi-Agent RAG Systems with DSPy and GEPA”...

Reposted by Lakshya A Agrawal

Tom Dörr @tom-doerr.bsky.social · Sep 9

optimizes prompts and code using AI-driven reflection and evolution

Screenshot of the repository

Reposted by Lakshya A Agrawal

Tom Dörr @tom-doerr.bsky.social · Sep 9

https://github.com/gepa-ai/gepa

Reposted by Lakshya A Agrawal

LLMs @llms.activitypub.awakari.com.ap.brid.gy · Sep 9

gepa 0.0.11 A framework for optimizing textual system components (AI prompts, code snippets, etc.) using LLM-based reflection and Pareto-efficient evolutionary search.

Origin | Interest | Match

Client Challenge

Reposted by Lakshya A Agrawal

Awakari @bluesky.awakari.com · Sep 9

gepa 0.0.11 A framework for optimizing textual system components (AI prompts, code snippets, etc.) using LLM-based reflection and Pareto-efficient evolutionary search.

Interest | Match | Feed

Reposted by Lakshya A Agrawal

Zeta Alpha @zeta-alpha.bsky.social · Sep 8

We’ll also cover new releases like EmbeddingGemma and the research shaping the field, including OpenAI’s “Why Language Models Hallucinate”, DeepMind’s “Theoretical Limitations of Embedding-Based Retrieval”, and recent work such as GEPA, BrowseComp-Plus, and Universal Deep Research.

See you there!

Reposted by Lakshya A Agrawal

LLMs @llms.activitypub.awakari.com.ap.brid.gy · Sep 4

DSPy and GEPA: Underrated Power Tools for AI Engineering Continue reading on Medium »

#large-language-models #technology #chatgpt #machine-learning #tech-companies

Origin | Interest | Match

DSPy and GEPA: Underrated Power Tools for AI Engineering

While the entire industry chases the next AI hype, tools that truly transform engineering practices often get lost in the noise. DSPy and its optimizer GEPA are prime examples — they’re quietly…

ai-engineering-trend.medium.com