| Website | http://prithvirajva.com |
I've been spending a lot of time thinking about #ChatGPT and academic assessment and why I am not so concerned about it (at the college level). It's not that ChatGPT is bad for education. It's that our education system is built upon a basis wherein we do assessment wrong.
What @hoffman writes in his blog below comes close to how I think of things (thanks for saving me a lot of time writing my thoughts down!)
http://write.guyhoffman.com/why-i-dont-care-if-students-use-gpt
In an information-seeking conversation, a user may ask questions that are under-specified or unanswerable. An ideal agent would interact by initiating different response types according to the available knowledge sources. However, most current studies either fail to or artificially incorporate such agent-side initiative. This work presents InSCIt, a dataset for Information-Seeking Conversations with mixed-initiative Interactions. It contains 4.7K user-agent turns from 805 human-human conversations where the agent searches over Wikipedia and either directly answers, asks for clarification, or provides relevant information to address user queries. The data supports two subtasks, evidence passage identification and response generation, as well as a human evaluation protocol to assess model performance. We report results of two systems based on state-of-the-art models of conversational knowledge identification and open-domain question answering. Both systems significantly underperform humans, suggesting ample room for improvement in future studies.
I'm pretty optimistic that the LLM reliability / factualness issue can be fixed. The key is to use LLMs as a dialog interface and not as a store of knowledge. LLMs as the query layer between a human user an a knowledge graph with sources (which can be hybrid generated/curated).
“I'm pretty optimistic that the LLM reliability / factualness issue can be fixed. The key is to use LLMs as a dialog interface and not as a store of knowledge. LLMs as the query layer between a human user an a knowledge graph with sources (which can be hybrid generated/curated).”
the success of chatgpt has lead to investors thinking RLHF is magic (to some extent it is), but boy they are going to be disappointed when their portfolios realize its limitations
Transformers are robust reasoners, but frustratingly lack the ability for accurate math, navigation, & other easily coded tasks. In our new work "Behavior Cloned Transformers are Neurosymbolic Reasoners", we show you can have the best of both worlds. 1/3
🐦🔗: https://twitter.com/peterjansen_ai/status/1580686608566583296
More good news, our paper on teaching interactive language agents to use existing symbolic tools and APIs has been accepted to #EACL2023. See y'all in Dubrovnik, Croatia!!!
In this work, we explore techniques for augmenting interactive agents with information from symbolic modules, much like humans use tools like calculators and GPS systems to assist with arithmetic and navigation. We test our agent's abilities in text games -- challenging benchmarks for evaluating the multi-step reasoning abilities of game agents in grounded, language-based environments. Our experimental study indicates that injecting the actions from these symbolic modules into the action space of a behavior cloned transformer agent increases performance on four text game benchmarks that test arithmetic, navigation, sorting, and common sense reasoning by an average of 22%, allowing an agent to reach the highest possible performance on unseen games. This action injection technique is easily extended to new agents, environments, and symbolic modules.
Now accepted to #ICLR2023! Look forward to our talk on open source, efficient natural language RLHF algorithms at Kigali, Rwanda!!!
The secret to aligning LMs to human preferences is reinforcement learning. But Why&How is it used? Announcing
💻RL4LMs: library to train any @[email protected] LM w/ RL
https://github.com/allenai/RL4LMs
👾GRUE: benchmark of 6 NLP tasks+rewards
📈NLPO: new RL alg 4 LMs
🌐https://rl4lms.apps.allenai.org
🐦🔗: https://twitter.com/rajammanabrolu/status/1577690380161585152
Mark Riedl
