Skip to content

Function Vectors in Large Language Models (ICLR 2024)

active 2024-06-032026-05-21 (UTC)

Partial coverage15,542 / 19,214 hourly files (81%) · 2 absent upstream · 3,671 failed, retryable2024-06-022026-08-11 (UTC)— sampled evenly across the window, so rankings and trends hold; absolute counts scale up.
Events
85
Pushes
1
Pull requests
0
Issues
12
Stars
54
Forks
8

Activity over time

Daily event counts in the loaded window

Line chart, 718 days from 2024-06-03 to 2026-05-21. Pushes: 1 total, peak 1 in a day. Pull requests: 0 total, peak 0 in a day. Issues: 12 total, peak 3 in a day. Comments: 10 total, peak 3 in a day. Stars: 54 total, peak 2 in a day.

  • Pushes
  • Pull requests
  • Issues
  • Comments
  • Stars

Stars, PRs, issues and forks are under-captured in the later part of this window. GH Archive progressively stopped capturing non-push events during 2026 — −95% or worse by the end of the window. Every series here except Pushes fades for that reason, so a decline above reflects the archive, not this repository. Pushes stay reliable throughout, so read them, and the contributor counts derived from them, as the real signal. Data health has the measurements.

Top contributors

Pushes, PRs, issues, reviews and comments — stars and forks excluded, so this is contribution rather than popularity

ContributorContributionsPushesPRsComments
harisethuram7002
ericwtodd6104
carolina-museum3001
RadiantCrystal2000
stripathy-eightfold2002
HieuNT911000
ST1720011001
zzFestinaLente1000

Recent activity

Latest issues, pull requests and releases

  • Issue#22zzFestinaLente2025-05-15 05:04
    Question about Simplifying Last Token Activation Extraction
  • Issue comment#17ericwtodd2025-05-13 00:48
    Getting all the indirect effect zeros for SeaLLM-v3 instruct model
  • Issue comment#16ericwtodd2025-05-13 00:47
    Unable to compute indirect effects in newer models such as LLama3, Qwen2
  • Issue comment#22ericwtodd2025-05-13 00:28
    Question about Simplifying Last Token Activation Extraction
  • Issue#21carolina-museum2025-04-20 17:57
    Clarification about attention head and function vector dimensions
  • Issue comment#21carolina-museum2025-04-20 17:57
    Clarification about attention head and function vector dimensions
  • Issue#21carolina-museum2025-04-17 00:13
    Clarification about attention head and function vector dimensions
  • Issue#15ericwtodd2025-03-14 21:36
    Does FV hold for mutli-token transformation tasks?
  • Issue comment#18ericwtodd2025-03-14 21:25
    gather_attn_activations() gives OOM after several iteration - but it seems not nowhere to accumulate GPU memory that could lead to the problem
  • Issue#17RadiantCrystal2025-01-19 21:22
    Getting all the indirect effect zeros for SeaLLM-v3 instruct model
  • Issue#16RadiantCrystal2025-01-19 13:35
    Unable to compute indirect effects in newer models such as LLama3, Qwen2
  • Issue#14harisethuram2024-08-12 04:51
    Gemma 2 / Mistral 7b v.1 issues
  • Issue#13harisethuram2024-07-26 21:55
    Creating top_heads list in compute_universal_function_vector (extract_utils.py) for new models
  • Issue comment#13harisethuram2024-07-26 21:55
    Creating top_heads list in compute_universal_function_vector (extract_utils.py) for new models
  • Issue#13harisethuram2024-07-26 21:55
    Creating top_heads list in compute_universal_function_vector (extract_utils.py) for new models
  • Issue#13harisethuram2024-07-26 21:52
    Creating top_heads list in compute_universal_function_vector (extract_utils.py) for new models
  • Issue#13harisethuram2024-07-21 18:01
    Creating top_heads list in compute_universal_function_vector (extract_utils.py) for new models
  • Issue comment#1harisethuram2024-07-21 17:41
    Where did "top_heads" come from in src/utils/extract_utils.py?
  • Issue#12HieuNT912024-06-04 03:04
    Performance metric used in the paper
  • Issue comment#10ST1720012024-06-03 03:37
    Computation of top_heads in compute_univeral_function_vector function.
  • Issue comment#10stripathy-eightfold2024-06-03 03:33
    Computation of top_heads in compute_univeral_function_vector function.
  • Issue comment#10stripathy-eightfold2024-06-03 03:28
    Computation of top_heads in compute_univeral_function_vector function.

Totals cover only the window loaded into ClickHouse and count events, not GitHub's lifetime totals — 54 stars here means stars gained during the window, not the repo's star count.