abductive-reasoning

The paper introduces the PIJ benchmark for evaluating large language models on criminal profiling tasks from incomplete evidence, highlighting performance gaps and biases in inferential reasoning.

This paper examines the alignment of thinking effort between humans and large reasoning models in abductive reasoning, finding evidence of shared effort and similar errors, and demonstrates that decoding methods increase this alignment.

This paper formalizes and measures the 'jump' in large language models, where they abandon default completions for correct ones, and finds that LLMs successfully jump in experimental trials.

The paper introduces CEDAR-GRPO, a process-aware reinforcement learning framework that enhances abductive reasoning in large language models by combining final-answer correctness with abductive rewards, demonstrating transferable improvements across various unseen tasks.

Introduces MoCA (Implicit Social Context Analysis), a new task and benchmark for modeling implicit social scenarios across affection, intent, and stance, along with a Conflict-Driven Abductive Reasoning (CoDAR) framework. Experiments show state-of-the-art multimodal LLMs struggle on this task, while CoDAR improves performance but still lags behind human reasoning.

This paper introduces Alien Abduction, an interactive game to probe how LLMs acquire evidence, update hypotheses, and decide when to stop during abductive reasoning. It finds that models perform better with upfront evidence and with oracle-provided examples than with self-selected queries, revealing deficiencies in active information acquisition.

This paper introduces NeurOWL, a neuro-symbolic framework using LLMs and ontology embeddings to perform subsumption verification and abduction on incomplete OWL ontologies, enabling reasoning when axioms are missing.

The paper introduces Elenchos, a generative evaluation framework for abductive reasoning in LLMs, where models must infer hidden rule changes from behavioral differences under black-box access. It finds a detection-attribution dissociation: models detect alterations but struggle to identify the specific mutations, especially under interacting mutations.

This paper presents a new paraconsistent AGM-like abductive expansion operation that can assimilate contradictory explanatory hypotheses without trivialization, based on the paraconsistent logic RCbr. It is the first operation of its kind in the AGM literature.

HypoAgent is an agentic framework for interactive abductive hypothesis generation over knowledge graphs, integrating three agents to handle evolving user intents and fine-grained diagnosis, achieving state-of-the-art performance.