REVIEW 6 cited by
CLadder: Assessing Causal Reasoning in Language Models
Not yet reviewed by Pith; the record is open.
This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.
SPECIMEN: schema-true, not a live event
T0 review · schema-true
One-sentence machine reading of the paper's core claim.
pith:XXXXXXXX · record.json · timestamp
read the original abstract
The ability to perform causal reasoning is widely considered a core feature of intelligence. In this work, we investigate whether large language models (LLMs) can coherently reason about causality. Much of the existing work in natural language processing (NLP) focuses on evaluating commonsense causal reasoning in LLMs, thus failing to assess whether a model can perform causal inference in accordance with a set of well-defined formal rules. To address this, we propose a new NLP task, causal inference in natural language, inspired by the "causal inference engine" postulated by Judea Pearl et al. We compose a large dataset, CLadder, with 10K samples: based on a collection of causal graphs and queries (associational, interventional, and counterfactual), we obtain symbolic questions and ground-truth answers, through an oracle causal inference engine. These are then translated into natural language. We evaluate multiple LLMs on our dataset, and we introduce and evaluate a bespoke chain-of-thought prompting strategy, CausalCoT. We show that our task is highly challenging for LLMs, and we conduct an in-depth analysis to gain deeper insights into the causal reasoning abilities of LLMs. Our data is open-sourced at https://huggingface.co/datasets/causalNLP/cladder, and our code can be found at https://github.com/causalNLP/cladder.
Forward citations
Cited by 6 Pith papers
-
STOCKTAKE: Measuring the Gap Between Perception and Action in LLM Agents with a Fair Oracle
Frontier LLM agents detect hidden supply-chain stress almost equally well (84–88% of episodes) but vary from skill 0.62 to −0.23, with two of four models acting worse than ignoring symptoms.
-
Reasoning Consensus: Structural Ensembling of LLM Reasoning via Weighted DAG Aggregation
Combining multiple LLMs' reasoning traces into weighted DAGs gives an auditable consensus graph that matches self-consistency and modestly improves on majority voting.
-
What General Intelligence Requires: Non-Reducible Constraints Across Levels of Description
General intelligence requires mutually non-reducible constraints across computational, biological, cultural, institutional, and incentive-theoretic levels, so no single architectural advance or scaling alone can produce AGI.
-
CausalFlip: A Benchmark for LLM Causal Judgment Beyond Semantic Matching
A new benchmark and training strategy show LLMs trained to internalize causal reasoning steps are less fooled by semantically similar, label-flipped questions than models using explicit chain-of-thought.
-
CausalAbstain: Enhancing Multilingual LLMs with Causal Reasoning for Trustworthy Abstention
CausalAbstain filters multilingual self-feedback by comparing how much it changes the model's abstention decision, improving abstention accuracy over baselines on two benchmarks.
-
Do Large Language Models Reason Causally Like Us? Even Better?
GPT-4o, Gemini-Pro, and Claude show less associative bias than human reasoners on collider graphs, making them more normatively aligned in likelihood judgments, but none fully demonstrates explaining away.
Discussion (0). Sign in to comment.