REVIEW 12 cited by
Does Fine-Tuning LLMs on New Knowledge Encourage Hallucinations?
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
Does Fine-Tuning LLMs on New Knowledge Encourage Hallucinations?
read the original abstract
When large language models are aligned via supervised fine-tuning, they may encounter new factual information that was not acquired through pre-training. It is often conjectured that this can teach the model the behavior of hallucinating factually incorrect responses, as the model is trained to generate facts that are not grounded in its pre-existing knowledge. In this work, we study the impact of such exposure to new knowledge on the capability of the fine-tuned model to utilize its pre-existing knowledge. To this end, we design a controlled setup, focused on closed-book QA, where we vary the proportion of the fine-tuning examples that introduce new knowledge. We demonstrate that large language models struggle to acquire new factual knowledge through fine-tuning, as fine-tuning examples that introduce new knowledge are learned significantly slower than those consistent with the model's knowledge. However, we also find that as the examples with new knowledge are eventually learned, they linearly increase the model's tendency to hallucinate. Taken together, our results highlight the risk in introducing new factual knowledge through fine-tuning, and support the view that large language models mostly acquire factual knowledge through pre-training, whereas fine-tuning teaches them to use it more efficiently.
Forward citations
Cited by 12 Pith papers
-
Reliability Scales Inversely: Hallucinations Snowball Faster in Bigger Language Models
Hallucination in LLMs is driven by an oracle-invisible “decoding risk” term that grows with scale and causally compounds errors within a response.
-
EHRAG: Bridging Semantic Gaps in Lightweight GraphRAG via Hybrid Hypergraph Construction and Retrieval
EHRAG constructs structural hyperedges from sentence co-occurrence and semantic hyperedges from entity embedding clusters, then applies hybrid diffusion plus topic-aware PPR to retrieve top-k documents, outperforming ...
-
Why Fine-Tuning Encourages Hallucinations and How to Fix It
Supervised fine-tuning increases LLM hallucinations via interference among overlapping semantic representations; self-distillation mitigates this by regularizing output-distribution drift while freezing parameters pre...
-
Understanding New-Knowledge-Induced Factual Hallucinations in LLMs: Analysis and Interpretation
Fine-tuning on new knowledge induces propagating hallucinations in LLMs by weakening attention to key entities, with mitigation via reintroducing known knowledge during later training stages.
-
Data Quality over Capacity: Internalizing Documents into LoRA Adapters for Closed-Book QA
For closed-book QA via LoRA adapters, once capacity is sufficient, training-data quality—especially answer brevity—outperforms rank, learning rate, and architecture changes, lifting 15-doc accuracy from 57.7% to 85.7%.
-
Semantic Layers for Reliable LLM-Powered Data Analytics: A Paired Benchmark of Accuracy and Hallucination Across Three Frontier Models
A paired benchmark demonstrates that providing an explicit semantic layer document improves LLM accuracy on text-to-SQL tasks by 17-23 percentage points and eliminates meaningful differences between frontier models.
-
REFLEX: Self-Refining Explainable Fact-Checking via Verdict-Anchored Style Control
REFLEX improves explainable fact-checking by using verdict-anchored style control and self-disagreement signals to disentangle fact from style in LLM outputs, achieving SOTA results with minimal self-refined samples.
-
REFLEX: Self-Refining Explainable Fact-Checking via Verdict-Anchored Style Control
By steering activations along directions learned from base-vs-fine-tuned disagreements, REFLEX reports 64.99 macro-F1 on RAW-FC fact-checking with only ~465 self-refined contrastive samples and no retrieval.
-
Reinforcement Learning Improves Traversal of Parametric Knowledge in LLMs
Reasoning LLMs seem better at retrieving hierarchical facts not because they know more but because they navigate better; the key supporting RL experiment is missing from the paper.
-
A Survey on Hallucination in Large Language Models: Principles, Taxonomy, Challenges, and Open Questions
The paper surveys hallucination in LLMs with an innovative taxonomy, factors, detection methods, benchmarks, mitigation strategies, and open research directions.
-
Position: Agent Should Invoke External Tools ONLY When Epistemically Necessary
Agents should invoke external tools only when epistemically necessary, per the introduced Theory of Agent framework that frames tool use as a decision under uncertainty.
-
Cultivating Machine Intelligence: The OMEGA Shift from Top-Down Optimization to Autopoietic Cognitive Ecologies
The paper introduces the RECLAIM framework and OMEGA shift as a transition from top-down optimization to autopoietic cognitive ecologies for cultivating machine intelligence.
discussion (0)
Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.