issues
search
QpxDesign
/
TREC-iKAT
An attempt of the iKAT (Interactive Knowledge Assistance Track) Track of the 2023 run of the NIST's TREC.
2
stars
1
forks
source link
issues
Newest
Newest
Most commented
Recently updated
Oldest
Least commented
Least recently updated
passage scores not indicator of which passages we're using, since some passages that were thrown out by TF-IDF reliability check or ChatGPT relevancy check may still have higher pyserni scores
#36
QpxDesign
closed
1 year ago
0
combined_passage_summaries length exceeds 250 spacy tokens
#35
QpxDesign
closed
1 year ago
0
Conduct run without TF-IDF reliability check
#34
QpxDesign
closed
1 year ago
0
double generating passage summaries, drastically diminishing performance
#33
QpxDesign
closed
1 year ago
0
put passages we used (ones marked with "used":True) at top of passage_provenance in run json for easier readability
#32
QpxDesign
closed
1 year ago
0
unclear whether we should use only the first topic in our submission or if we can choose which topic to submit
#31
QpxDesign
closed
1 year ago
1
use two or n shot approach
#30
QpxDesign
closed
1 year ago
1
ptkb selection could use some work
#29
QpxDesign
closed
1 year ago
0
weird 'meta' response
#28
QpxDesign
closed
1 year ago
0
implement n shot support (rather than just one or zero being hardcoded)
#27
QpxDesign
closed
1 year ago
0
also report reliability score from passage classifier
#26
QpxDesign
closed
1 year ago
0
answer_question_from_passage doesn't take into account conversation history
#25
QpxDesign
closed
1 year ago
0
switch from basic LucerneSearcher to a HybridSearcher/DenseSearcher for better results when finding passages
#24
QpxDesign
closed
1 year ago
1
if passage selection is bad it will generate some nonsense
#23
QpxDesign
closed
1 year ago
1
Incorrect trimming of 'listing' response in prevent_trail_off
#22
QpxDesign
closed
1 year ago
0
dialogue history for LLAMA is based on responses given in the test json
#21
QpxDesign
closed
1 year ago
1
straight up didn't respond to question
#20
QpxDesign
closed
1 year ago
1
trailing off after colon
#19
QpxDesign
closed
1 year ago
1
if passage doesn't contain enough/right info it will refuse to answer question
#18
QpxDesign
closed
1 year ago
1
responses mention fact they are based off passages, not sure if thats desired behavior
#17
QpxDesign
closed
1 year ago
1
incorporated passage summaries & awnsering questions based on passage…
#16
QpxDesign
closed
1 year ago
0
unsure of proper JSON structure for submisson
#15
QpxDesign
closed
1 year ago
1
pulling is what i do
#14
QpxDesign
closed
1 year ago
0
can't use 70B param llama2 - doesn't have option for n_ctx
#13
QpxDesign
closed
1 year ago
1
llama-cpp-python not saturating GPU
#12
QpxDesign
closed
1 year ago
1
retrieved passages arent being used in llm to generate responses
#11
QpxDesign
closed
1 year ago
1
llm responses cut off due to hard 250 token limit
#10
QpxDesign
closed
1 year ago
0
pull go brrrr
#9
QpxDesign
closed
1 year ago
0
tweaked ptkb threshold & improved status indicators for full-run
#8
QpxDesign
closed
1 year ago
0
Quinn
#7
QpxDesign
closed
1 year ago
0
added clueweb support into get-passages and readme
#6
QpxDesign
closed
1 year ago
0
trim ptkbs based on threshold
#5
QpxDesign
closed
1 year ago
0
Quinn
#4
QpxDesign
closed
1 year ago
0
can i merge pls
#3
QpxDesign
closed
1 year ago
1
Added Llama2, changed file structure
#2
QpxDesign
closed
1 year ago
0
Merging ptkb_solution logic and file structure changes
#1
QpxDesign
closed
1 year ago
0