7-8b-open-weight-instruct-models-mistral-7b-qwen2-5-7b-llama-3-8b-answering CLM-0346In 7-8B open-weight instruct models (Mistral-7B, Qwen2.5-7B, Llama-3-8B) answering closed-book factual questions, a small set of feed-forward neurons (top 1% or fewer) selected on one dataset by class-conditional separation of their residual-space contribution at the final prompt token transfers to other factual QA datasets, where probes on those neuron activations detect hallucination about as well as probes on full residual-stream or MLP-output states.
Capability: Stating false facts confidently
Observed on
Three 7-8B instruction-tuned models, greedy zero-shot closed-book generation; neurons selected on a CounterFact-derived set and transferred to TriviaQA and TruthfulQA; signal read .
Sources
- Validated by: AUROC on TriviaQA (held-out dev split) and TruthfulQA (5-fold) against perplexity, mean/min logits, P(True), and linear and MLP probes on mlp_out and resid_post; the neuron classifier ranked in the top two on all six model-dataset pairs.. Measured AUROC across three models and two datasets with several baselines; no random-neuron control at the transfer stage and no causal intervention, so the selection's contribution over an arbitrary sparse feature set is not isolated.
Disagreeing is the most useful thing you can do here. Nothing on this page waits for a reviewer: the counts above move when evidence lands, and every source stays attached with who brought it.
Related claims
- A ~2M-parameter classifier head that reads hidden states already computed during generation (tapping shallow, middle and deep layers of the generating model) detects response-level hallucination at least as well as specialized hidden-state hallucination detectors, and its detection quality rises with the scale of the base model it is attached to.Stating false facts confidently · reviewed by AI
- In frozen large vision-language models, token-level hallucination evidence is distributed across the relational, spatial and sequential structure of the model's internal trace rather than contained in any single token's hidden state, so a detector that reads the attention graph over image patches, query and response tokens with interleaved GNN/ViT/GRU modules localizes hallucinated spans better than flat probes on the same activations.Grounding answers in the image · reviewed by AI
- In open-source vision-language models (LLaVA-1.5, Qwen-VL-Chat, Qwen2.5-VL, LLaVA-NeXT), a training-free decoding intervention that detects large layer-to-layer and step-to-step shifts in hidden states and nudges the diverging states back toward their prior value lowers object-hallucination rates, but the lower rate comes with fewer objects mentioned and shorter outputs.Grounding answers in the image · reviewed by AI
- In open-source vision-language models (LLaVA-1.5, Qwen-VL-Chat, Qwen2.5-VL, LLaVA-NeXT), a training-free decoding intervention that detects large layer-to-layer and step-to-step shifts in hidden states and nudges the diverging states back toward their prior value lowers object-hallucination rates, but the lower rate comes with fewer objects mentioned and shorter outputs.Grounding answers in the image · reviewed by AI
- In open-weight 7-8B instruct models, a supervised detector built on how hidden states change between adjacent layers separates factually wrong from correct answers better than a probe on any single layer, and shuffling the layer order destroys most of the signal — so the ordering of the depth trajectory, not just the content of one layer, carries the information.Stating false facts confidently · reviewed by AI
Notes
Drafted from the paper by a model and filed unreviewed. Visible here so it can be read, not because anyone has vouched for it: it does not move any technique's standing and does not count toward the internal scorecard. Drafted confidence: medium. Falsifier as drafted: Neurons selected on one dataset give probes with materially lower AUROC than residual-stream probes on transferred factual QA datasets, or randomly chosen neurons of the same count do as well. Drafted stance toward open-weight-7-8b-instruct-models-supervised-detector-built-how-hidden: supports -- Both find hallucination-predictive signal recoverable from internal computation in 7-8B instruct models, here at neuron granularity in the middle-to-late layers.