MAP-Law dynamically controls retrieval depth in legal AI by computing element coverage, evidence coverage, and marginal gain on a joint node graph, reaching 0.86 element coverage with 58% fewer rounds than fixed baselines on 50 labor-law cases.
Large legal fictions: Profiling legal hallucinations in large language models
7 Pith papers cite this work, alongside 14 external citations. Polarity classification is still indexing.
7
Pith papers citing it
14
external citations · Pith
abstract
Do large language models (LLMs) know the law? These models are increasingly being used to augment legal practice, education, and research, yet their revolutionary potential is threatened by the presence of hallucinations -- textual output that is not consistent with legal facts. We present the first systematic evidence of these hallucinations, documenting LLMs' varying performance across jurisdictions, courts, time periods, and cases. Our work makes four key contributions. First, we develop a typology of legal hallucinations, providing a conceptual framework for future research in this area. Second, we find that legal hallucinations are alarmingly prevalent, occurring between 58% of the time with ChatGPT 4 and 88% with Llama 2, when these models are asked specific, verifiable questions about random federal court cases. Third, we illustrate that LLMs often fail to correct a user's incorrect legal assumptions in a contra-factual question setup. Fourth, we provide evidence that LLMs cannot always predict, or do not always know, when they are producing legal hallucinations. Taken together, our findings caution against the rapid and unsupervised integration of popular LLMs into legal tasks. Even experienced lawyers must remain wary of legal hallucinations, and the risks are highest for those who stand to benefit from LLMs the most -- pro se litigants or those without access to traditional legal resources.
citation-role summary
background 1
citation-polarity summary
roles
background 1polarities
background 1representative citing papers
A learned weighted sum of intermediate-layer activations compresses an instruction prompt into a single patch vector that, injected at an early layer, recovers task accuracy within ~2% of the full prompt.
Creates TF-RefusalBench to quantify over-alignment in LLMs on criminal-law tasks across four languages and shows abliteration mitigates refusals with little performance loss.
Steering LLM residual streams with random sparse vectors creates detectable self-recognition fingerprints that enable over 98% accurate attribution of generated text to specific models without degrading output quality.
Authority-style prefixes increase refusal rates 2-20x in small on-prem LLMs on criminal legal prompts while a jailbreak prefix shows mixed effects.
The paper introduces a taxonomy of AI safety for LLMs organized into Trustworthy AI, Responsible AI, and Safe AI perspectives, accompanied by a review of state-of-the-art methods, challenges, and future directions.
A framework detects LLM anomalies including hallucinations, jailbreaks, and backdoors by forensic inspection of layer-wise hidden state patterns, reporting over 95% accuracy with minimal computational overhead.
citing papers explorer
-
MAP-Law: Coverage-Driven Retrieval Control for Multi-Turn Legal Consultation
MAP-Law dynamically controls retrieval depth in legal AI by computing element coverage, evidence coverage, and marginal gain on a joint node graph, reaching 0.86 element coverage with 58% fewer rounds than fixed baselines on 50 labor-law cases.
-
Prompt Compression via Activation Aggregation
A learned weighted sum of intermediate-layer activations compresses an instruction prompt into a single patch vector that, injected at an early layer, recovers task accuracy within ~2% of the full prompt.
-
Measuring & Mitigating Over-Alignment for LLMs in Multilingual Criminal Law Courts
Creates TF-RefusalBench to quantify over-alignment in LLMs on criminal-law tasks across four languages and shows abliteration mitigates refusals with little performance loss.
-
LLM Self-Recognition: Steering and Retrieving Activation Signatures
Steering LLM residual streams with random sparse vectors creates detectable self-recognition fingerprints that enable over 98% accurate attribution of generated text to specific models without degrading output quality.
-
LLMs Prompted for Legal Context Object More: Overrefusal from Small On-Premises LLMs in Criminal Legal Context
Authority-style prefixes increase refusal rates 2-20x in small on-prem LLMs on criminal legal prompts while a jailbreak prefix shows mixed effects.
-
AI Safety Landscape for Large Language Models: Taxonomy, State-of-the-art, and Future Directions
The paper introduces a taxonomy of AI safety for LLMs organized into Trustworthy AI, Responsible AI, and Safe AI perspectives, accompanied by a review of state-of-the-art methods, challenges, and future directions.
-
Exposing the Ghost in the Transformer: Abnormal Detection for Large Language Models via Hidden State Forensics
A framework detects LLM anomalies including hallucinations, jailbreaks, and backdoors by forensic inspection of layer-wise hidden state patterns, reporting over 95% accuracy with minimal computational overhead.