Issues

• [Bug Report] success_rate_ci returns wrong Clopper-Pearson bounds at high success rates or large n (binomial CDF underflows)

  TransformerLensOrg/TransformerLens#1926 · Filter by author akshpatel0910 opened on Oct 21, 2026

• [Bug Report] Layer-0 stack_head_results / stack_neuron_results are float32 while the other layers follow the cache

  TransformerLensOrg/TransformerLens#1924 · Filter by author avgvi opened on Oct 21, 2026

• [Bug Report] test_prompt with several answers reports the pad token's rank when the tokenizer pads on the left

  TransformerLensOrg/TransformerLens#1922 · Filter by author avgvi opened on Oct 21, 2026

• [Bug Report] FactoredMatrix SVD, composition scores and SVDInterpreter fail on bf16/fp16 bridges

  TransformerLensOrg/TransformerLens#1918 · Filter by author tivris opened on Oct 20, 2026
  2

  

• [Proposal] Neuron bases for interchange interventions

  TransformerLensOrg/TransformerLens#1917 · Filter by author dipo101 opened on Oct 20, 2026
  1

  

• [Bug Report] Granite MoE Hybrid grouped experts produce incorrect or unstable outputs on MPS

  bugcomplexity-simpleTransformerBridge
  Bug

  TransformerLensOrg/TransformerLens#1886 · Filter by author janmenjayap opened on Oct 12, 2026
  

• [Proposal] Natural Language Autoencoders: explain and reconstruct Qwen activations through TransformerBridge

  Community Feedback RequestedResearch Reproductionseen_by_maintainersTransformerBridge
  TransformerLensOrg/TransformerLens#1884 · Filter by author janmenjayap opened on Oct 11, 2026
  1

• [Proposal] Contextual Decomposition: Nonlinear Source Attribution and Dataset-Wide Ranking on GPT-2

  Community Feedback RequestedResearch Reproductionseen_by_maintainers
  TransformerLensOrg/TransformerLens#1883 · Filter by author janmenjayap opened on Oct 11, 2026
  1

• [Proposal] Representation Geometry: causal inner products, concept directions, and categorical diagnostics

  Community Feedback Requestedenhancementseen_by_maintainersTransformerBridge
  TransformerLensOrg/TransformerLens#1880 · Filter by author janmenjayap opened on Oct 10, 2026
  1

• [Proposal] RoBERTa masked-LM adapter for TransformerBridge

  complexity-moderatenew-architectureTransformerBridge
  TransformerLensOrg/TransformerLens#1870 · Filter by author Canonik opened on Oct 8, 2026
  1

  

• [Bug Report] _BLOCK_LIST_ATTRS hardcoded name list silently drops Raven's blocks from composition-score / head-label analysis

  bugcomplexity-moderateTransformerBridge
  TransformerLensOrg/TransformerLens#1791 · Filter by author LightWork666 opened on Sep 18, 2026
  2