Akshara Prabhakar
@aksh_555
Followers
401
Following
1K
Media
7
Statuses
70
applied scientist @SFResearch | prev @princeton_nlp, @surathkal_nitk
bay area
Joined October 2019
🤖 NEW PAPER 🤖. Chain-of-thought reasoning (CoT) can dramatically improve LLM performance. Q: But what *type* of reasoning do LLMs use when performing CoT? Is it genuine reasoning, or is it driven by shallow heuristics like memorization?. A: Both!. 🔗 1/n
7
48
315
RT @quantumbytz: Salesforce AI Introduces CRMArena-Pro: The First Multi-Turn and Enterprise-Grade Benchmark for LLM Agents.#AI #MachineLear….
0
2
0
RT @SFResearch: 🚨 Introducing CRMArena-Pro: The first multi-turn, enterprise-grade benchmark for LLM agents. ✍️Blog: .
0
26
0
RT @salesforce: .@SFResearch’s new series “AI Research Lab - Explained” just dropped!. First up? See how we fine-tune specialized models to….
0
5
0
RT @silviocinguetta: Enterprise General Intelligence (EGI) won't require bigger models—it will demand better data! Our recent research demo….
0
4
0
🚀 Just dropped APIGen-MT-5k — 5K high-quality multi-turn agent interactions, generated with our APIGen-MT framework! Built for training & evaluating AI agents.
Introducing APIGen-MT: Our agentic pipeline for multi-turn synthetic data generation that produces high-quality training data for tuning AI agents! Try our open-sourced dataset today!. 📊 Paper: .🤗 Dataset: We used APIGen-MT to
1
2
14
RT @AndrewLampinen: New preprint! In “Naturalistic Computational Cognitive Science: Towards generalizable models and theories that capture….
0
81
0
It was a great experience interacting with the @buZZrobot community 😃, thanks for the invite @sopharicks!. Talk link:
Join us on Feb 20 for a talk on Generalization vs. Memorization in LLMs. @aksh_555 from @SFResearch will dive into how Chain-of-Thought prompting impacts reasoning—does it truly enhance logic, or is it just smart memorization?.
0
0
11
RT @SFResearch: 🤖 Fresh from #NeurIPS2024: Our AI research scientist Akshara Prabhakar @aksh_555 discusses our demo of xLAM's specialized a….
0
4
0
RT @SFResearch: 🇨🇦🇨🇦🇨🇦 Welcome to Vancouver! 🇨🇦🇨🇦🇨🇦.13 Paper links below! 👇. The @Salesforce AI Research team brought a baker's dozen AI Re….
0
7
0
RT @iscreamnearby: It actually reminds me of the multi-source domain adaptation work - where knowing domain index during training makes the….
0
1
0
Work done with the amazing team: Yuanzhi Li, @karthik_r_n, @ShamKakade6, and guided by Samy Jelassi and @EranMalach. To appear in COLING 2025 Industry track!. 6/6.
0
0
1
🤔 Limitations.- Scaling to more than two tasks is still a challenge.- Initial experiments show scope for further improvements!. 5/n.
1
0
1
🌟 We explore 4 practical settings.- solving hard math problems with code.- domain specific question-answering bots.- comprehension on technical documents.- robustness to prompt formats. 4/n
1
0
1
🔄 The traditional approach? Training over a mixture of the two datasets. However, concatenating LoRAs performs better than this and other merging methods (TIES, DARE, LoRAHub) 🎉. 3/n
1
0
0
🤖 Imagine a question-answering bot for a medical textbook without RAG. You can train a model for next-token prediction on the textbook, and use an instruction-following model (trained on Alpaca for eg.) and compose them!. 2/n.
1
0
0