Here they are with a brief summary of what I got from each:
Collision for Hash Functions MD4, MD5 HAVAL-128 and RIPEMDXiaoyun Wang, Dengguo Feng, Xuejia Lai and Hongbo Yu
This is the original paper listing out some collisions for each of these functions. This must have been quite a blockbuster at the time.
Cryptanalysis of the Hash Functions MD4 and RIPEMDXiaoyun Wang, Xuejia Lai, Dengguo Feng, Hui Chen, and Xiuyuan Yu
This article details the attack that was used to generate the collisions of the previous paper and should be all you need to write a collision generating script for MD4 and RIPEMD.
How to Break MD5 and Other Hash Functions
Xiaoyun Wang and Hongbo Yu
MD5 is slightly harder to break than MD4 requiring 2 blocks and more muli-step message modifications. This article details the method used to generate MD5 collisions in the first.
Searching for Differential Paths in MD4Martin Schälffer and Elisabeth Oswald
More detail on how the attacks work with a good description of how paths are calculated and an algorithm for finding them. Also contains a new path with fewer stage 2 required requirements.
Improved Collision Attack on MD5
Yu Sasaki, Yusuke Naito, Noboru Kunihiro and Kazuo Ohtahttps://s3-eu-west-1.amazonaws.com/md5collisions/ImprovedCollisionAttackonMD5.pdf
The paper where I finally understood how the correction of second round collisions worked
Improved Collision Attack on MD4
Yusuke Naito, Yu Sasaki, Noboru Kunihiro, and Kazuo Ohta
Some corrections to the Wang collision on MD4 speeds things up with good explanation.
Automatic Search of Differential Path in MD4
Pierre-Alain Fouque, Gaëtan Leurent, Phong Nguyen
New Message Difference for MD4
Yu Sasaki, Lei Wang, Kazuo Ohta and Noboru Kunihiro
The best path I know of with a totally different message difference and explanation of the local collisions underlying the collisions.
Herding Hash Functions and the Nostradamus Attack
John Kelsey and Tadayoshi Kohno