technion-cs-nlp

bias-probing

Debiasing Methods in Natural Language Understanding Make Bias More Accessible: Code and Data

BiologicalTokenizers

Effect of tokenization on transformers for biological sequence

gemini-technion

A Technion-branded theme for the Gemini LaTeX beamerposter theme

gender_internal

hallucination-mitigation

Individual-Neurons-Pitfalls

irm-for-nli

llm-arithmetic-heuristics

LLMsKnow

ReFACT

technion-cs-nlp/LLMsKnow

technion-cs-nlp/hallucination-mitigation

technion-cs-nlp/BiologicalTokenizers

Effect of tokenization on transformers for biological sequence

technion-cs-nlp/llm-arithmetic-heuristics

technion-cs-nlp/bias-probing

Debiasing Methods in Natural Language Understanding Make Bias More Accessible: Code and Data

technion-cs-nlp/irm-for-nli

technion-cs-nlp/Individual-Neurons-Pitfalls

technion-cs-nlp/ReFACT

technion-cs-nlp/gemini-technion

A Technion-branded theme for the Gemini LaTeX beamerposter theme

technion-cs-nlp/gender_internal

technion-cs-nlp/ContraSim

A Similarity Measure Based on Contrastive Learning

technion-cs-nlp/Generative-NLI

An unbiased model for Natural Language Inference using generative models.

technion-cs-nlp/Information-Reflects-Memorization

technion-cs-nlp/BLIND

This git repo contains the all of the code required to run and reproduce the results from the paper BLIND: Bias Removal With No Demographics (ACL 23). For questions, write to orgad.hadas@cs.technion.ac.il.

technion-cs-nlp/idani

IDANI: Inference-time Domain Adaptation via Neuron-level Interventions

technion-cs-nlp/igbp_nonlinear-removal

This repository contains the code implementation for the Iterative Gradient-Based Projection (IGBP) algorithm proposed in the paper titled 'Shielded Representations: Protecting Sensitive Attributes Through Iterative Gradient-Based Projection'. Removing non-linear encoded concepts from neural representations and mitigate bias in NLP models."

technion-cs-nlp/metaphor-detection

technion-cs-nlp/DAFair

Mitigating social bias in language models without relying on deomgraphic labels

technion-cs-nlp/nlp4aja

NLP tools for processing Algerian Judeo-Arabic

technion-cs-nlp/Fast-Forward

Fast Forward the finetuning of LLMs

technion-cs-nlp/InstructedToBias

This repository is the code base for the paper: Instructed to Bias: Instruction-Tuned Language Models Exhibit Emergent Cognitive Bias (TACL).

technion-cs-nlp/REVS

REVS: Unlearning Sensitive Information in LLMs