🥳 "Logically Consistent Language Models via Neuro-Symbolic Integration" just accepted at #ICLR2025! We focus on instilling logical rules in LLMs with an efficient loss, leading to higher factuality & (self) consistency. How? 🧵 - ThreadSky

About ThreadSky

diadochus.bsky.social • 29 days ago

🥳 "Logically Consistent Language Models via Neuro-Symbolic Integration" just accepted at #ICLR2025!
We focus on instilling logical rules in LLMs with an efficient loss, leading to higher factuality & (self) consistency. How? 🧵

Comments

diadochus.bsky.social•29 days ago

Various background works focus on instilling single consistency rules, e.g. A and not A can't be both true (negation, Burns et al.), A true and A implies B, thus B true (modus ponens). Can we derive a general objective function that combines logical rules dynamically?

diadochus.bsky.social•29 days ago

Yes! We propose to leverage the Semantic Loss as a regularizer: it maximizes the likelihood of world (model) assignments satisfying any given logical rule. We thus include efficient solvers in the training pipeline to efficiently perform model counting on the LLM's own beliefs.

diadochus.bsky.social•29 days ago

Our method makes LLaMa's knowledge more consistent to any given knowledge graph, by seeing only a portion of it! It can transfer logical rules to similar or derived concepts. As proposed by @ekinakyurek.bsky.social et al., you can use a LLM-generated KB to reason over its knowledge.

diadochus.bsky.social•29 days ago

Finally, LOgically COnsistent (LoCo) LLaMas can outperform solver-based baselines and SFT! I thank @nolovedeeplearning.bsky.social and @looselycorrect.bsky.social for the guidance in realizing this project, get in touch or come to chat in Singapore!

https://arxiv.org/abs/2409.13724

Posting Rules

Be respectful to others
No spam or self-promotion
Stay on topic
Follow Bluesky's terms of service

Comments

Posting Rules

Reply